{"id":5287,"date":"2026-02-24T04:14:15","date_gmt":"2026-02-24T04:14:15","guid":{"rendered":"https:\/\/research.undetectable.ai\/?p=5287"},"modified":"2026-02-24T04:14:50","modified_gmt":"2026-02-24T04:14:50","slug":"i-tested-5-ai-image-detectors-these-are-the-best-in-2026","status":"publish","type":"post","link":"https:\/\/research.undetectable.ai\/pl\/przetestowalem-5-detektorow-obrazu-ai-te-sa-najlepsze-w-2026-roku\/","title":{"rendered":"Przetestowa\u0142em 5 wykrywaczy obraz\u00f3w AI. Te s\u0105 najlepsze w 2026 roku:"},"content":{"rendered":"

Prawdopodobnie widzia\u0142e\u015b w sieci wi\u0119cej zdj\u0119\u0107 wygenerowanych przez sztuczn\u0105 inteligencj\u0119, ni\u017c ci si\u0119 wydaje. <\/strong><\/p>\n\n\n\n

Czasami oczywiste jest, \u017ce obraz zosta\u0142 wygenerowany przez sztuczn\u0105 inteligencj\u0119, ale coraz trudniej jest to stwierdzi\u0107 w miar\u0119 ulepszania narz\u0119dzi do generowania obraz\u00f3w i wideo. Nowe narz\u0119dzia, takie jak Google Nano Banana Pro<\/a>i aktualizacje OpenAI Model obrazu ChatGPT<\/a> pozwalaj\u0105 u\u017cytkownikom szybko generowa\u0107 syntetyczne obrazy, kt\u00f3re odzwierciedlaj\u0105 rzeczywiste. Poprzednie badania wykaza\u0142y, \u017ce 85% Amerykan\u00f3w twierdzi, \u017ce Podr\u00f3bki podkopuj\u0105 zaufanie w sieci<\/a>.<\/p>\n\n\n\n

Wiele narz\u0119dzi twierdzi, \u017ce wykrywa obrazy generowane przez sztuczn\u0105 inteligencj\u0119, ale czy dzia\u0142aj\u0105? <\/strong><\/h2>\n\n\n\n

Przeprowadzi\u0142em 50 test\u00f3w wykrywania w pi\u0119ciu najpopularniejszych detektorach obraz\u00f3w AI i udokumentowa\u0142em wyniki. Nie tylko przedstawi\u0119 wszystkie dane w tym artykule i wyja\u015bni\u0119 je, ale tak\u017ce podam link do dokumentacji na ko\u0144cu.<\/p>\n\n\n\n

Do test\u00f3w u\u017cy\u0142em pi\u0119ciu detektor\u00f3w: TruthScan<\/a>, Sztuczna inteligencja czy nie<\/a>, Silnik celowniczy<\/a>, WasItAI<\/a>oraz Winston AI<\/a>. Przetestowa\u0142em te narz\u0119dzia, ka\u017c\u0105c ka\u017cdemu z nich przeanalizowa\u0107 dwa obrazy wygenerowane przez ChatGPT, sze\u015b\u0107 obraz\u00f3w wygenerowanych przez Nano Banana i dwa obrazy wygenerowane przez Midjourney. U\u017cy\u0142em wielu styl\u00f3w i technik podpowiedzi (kt\u00f3re szczeg\u00f3\u0142owo om\u00f3wi\u0119 p\u00f3\u017aniej, gdy b\u0119d\u0119 analizowa\u0142 ka\u017cdy analizowany obraz).<\/p>\n\n\n\n

Wszystkie pi\u0119\u0107 detektor\u00f3w mia\u0142o za zadanie wykry\u0107 obrazy stworzone przez sztuczn\u0105 inteligencj\u0119 w kategoriach takich jak oszustwo, dezinformacja, fotografia og\u00f3lna i deepfake. Nic dziwnego, \u017ce nie wszystkie detektory wypad\u0142y dobrze. TruthScan by\u0142 jedynym wykrywaczem, kt\u00f3ry konsekwentnie klasyfikowa\u0142 wszystkie przes\u0142ane przeze mnie tre\u015bci na poziomie 97% lub wy\u017cszym.<\/strong><\/p>\n\n\n\n\n\n\n\n\nWyniki testu wykrywacza obraz\u00f3w AI<\/title>\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .ai-detector-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .ai-detector-wrap h2 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.75rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .ai-detector-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .ai-detector-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.85rem;\n text-transform: uppercase;\n letter-spacing: 0.06em;\n padding: 16px 20px;\n text-align: center;\n border: none;\n }\n\n .ai-detector-table thead th:first-child {\n text-align: left;\n padding-left: 24px;\n }\n\n .ai-detector-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .ai-detector-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .ai-detector-table tbody td {\n padding: 14px 20px;\n text-align: center;\n font-size: 0.95rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .ai-detector-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .ai-detector-table tbody td:first-child {\n text-align: left;\n padding-left: 24px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n \/* Colored cells *\/\n .cell-green {\n background: #22c55e;\n color: #fff !important;\n font-weight: 700 !important;\n border-radius: 6px;\n padding: 6px 12px;\n display: inline-block;\n min-width: 32px;\n text-align: center;\n font-size: 0.9rem;\n }\n\n .cell-red {\n background: #ef4444;\n color: #fff !important;\n font-weight: 700 !important;\n border-radius: 6px;\n padding: 6px 12px;\n display: inline-block;\n min-width: 32px;\n text-align: center;\n font-size: 0.9rem;\n }\n\n \/* Accuracy badges *\/\n .badge-100 { background: #16a34a; color: #fff; }\n .badge-80 { background: #a3e635; color: #1a1a2e; }\n .badge-70 { background: #facc15; color: #1a1a2e; }\n .badge-60 { background: #fb923c; color: #fff; }\n .badge-30 { background: #ef4444; color: #fff; }\n\n .accuracy-badge {\n font-weight: 800;\n font-family: 'Outfit', sans-serif;\n font-size: 0.95rem;\n border-radius: 8px;\n padding: 6px 16px;\n display: inline-block;\n min-width: 60px;\n text-align: center;\n letter-spacing: 0.02em;\n }\n\n \/* Row rank indicator *\/\n .rank-icon {\n margin-right: 6px;\n font-size: 1rem;\n }\n\n \/* Asterisk note *\/\n .table-note {\n margin-top: 0.75rem;\n font-size: 0.8rem;\n color: #888;\n padding-left: 4px;\n }\n\n \/* Responsive *\/\n @media (max-width: 600px) {\n .ai-detector-table thead th,\n .ai-detector-table tbody td {\n padding: 10px 10px;\n font-size: 0.8rem;\n }\n .ai-detector-wrap h2 {\n font-size: 1.3rem;\n }\n }\n<\/style>\n<\/head>\n<body style=\"background:#f8f8fa; padding: 2rem;\">\n\n<div class=\"ai-detector-wrap\">\n <h2>Wyniki testu detektora obrazu AI<\/h2>\n\n <table class=\"ai-detector-table\">\n <thead>\n <tr>\n <th>Narz\u0119dzie do wykrywania<\/th>\n <th>Testy og\u00f3\u0142em<\/th>\n <th>Prawid\u0142owo wykryty<\/th>\n <th>Chybiony (Nieudany)<\/th>\n <th>Dok\u0142adno\u015b\u0107<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td><span class=\"rank-icon\">\ud83e\udd47<\/span>TruthScan<\/td>\n <td>10<\/td>\n <td><span class=\"cell-green\">10<\/span><\/td>\n <td><span class=\"cell-green\">0<\/span><\/td>\n <td><span class=\"accuracy-badge badge-100\">100%<\/span><\/td>\n <\/tr>\n <tr>\n <td><span class=\"rank-icon\">\ud83e\udd48<\/span>Sztuczna inteligencja czy nie<\/td>\n <td>10<\/td>\n <td><span class=\"cell-green\">8<\/span><\/td>\n <td><span class=\"cell-red\">2*<\/span><\/td>\n <td><span class=\"accuracy-badge badge-80\">80%<\/span><\/td>\n <\/tr>\n <tr>\n <td><span class=\"rank-icon\">\ud83e\udd49<\/span>Silnik celowniczy<\/td>\n <td>10<\/td>\n <td><span class=\"cell-green\">7<\/span><\/td>\n <td><span class=\"cell-red\">3<\/span><\/td>\n <td><span class=\"accuracy-badge badge-70\">70%<\/span><\/td>\n <\/tr>\n <tr>\n <td><span class=\"rank-icon\">   <\/span>WasItAI<\/td>\n <td>10<\/td>\n <td><span class=\"cell-green\">6<\/span><\/td>\n <td><span class=\"cell-red\">4<\/span><\/td>\n <td><span class=\"accuracy-badge badge-60\">60%<\/span><\/td>\n <\/tr>\n <tr>\n <td><span class=\"rank-icon\">   <\/span>Winston AI<\/td>\n <td>10<\/td>\n <td><span class=\"cell-green\">3<\/span><\/td>\n <td><span class=\"cell-red\">7<\/span><\/td>\n <td><span class=\"accuracy-badge badge-30\">30%<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n\n <p class=\"table-note\">* AI or Not oznaczy\u0142o jedn\u0105 generacj\u0119 AI jako 78%, jedn\u0105 jako 85% i jedn\u0105 jako 89%. Poniewa\u017c test wymaga co najmniej 90%, liczymy pierwsze dwa jako nieudane, ale traktujemy 89% jako prawie nieudany, st\u0105d gwiazdka.<\/p>\n<\/div>\n\n<\/body>\n<\/html>\n\n\n\n<p style=\"font-size:17px\">AI or Not r\u00f3wnie\u017c wypad\u0142 ca\u0142kiem dobrze. W rzeczywisto\u015bci, podczas gdy AI or Not sklasyfikowa\u0142 wiele element\u00f3w wygenerowanych przez AI poni\u017cej pewno\u015bci 90%, by\u0142 to drugi najdok\u0142adniejszy i najbardziej sp\u00f3jny detektor podczas moich test\u00f3w, za TruthScan. Nawet tam, gdzie AI or Not zawiod\u0142o, awarie te nie by\u0142y katastrofalne, ale wci\u0105\u017c jest miejsce na popraw\u0119. <\/p>\n\n\n\n<p style=\"font-size:17px\">Pozosta\u0142e testowane przeze mnie wykrywacze obraz\u00f3w AI radzi\u0142y sobie znacznie gorzej. Na przyk\u0142ad Sight Engine b\u0142\u0119dnie zaklasyfikowa\u0142 3 obrazy AI zwi\u0105zane z oszustwami jako autentyczne.<\/p>\n\n\n\n<p style=\"font-size:17px\">Teraz poka\u017c\u0119 ci ka\u017cdy obraz (z 10, kt\u00f3re wygenerowa\u0142em), wyja\u015bni\u0119, jak go stworzy\u0142em i poka\u017c\u0119, jak ka\u017cdy model go oceni\u0142.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#1. \"Man On a Ledge\" (Wygenerowane przez ChatGPT)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"683\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge-683x1024.jpg\" alt=\"\" class=\"wp-image-5289\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge-683x1024.jpg 683w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge-200x300.jpg 200w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge-768x1152.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge-8x12.jpg 8w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge.jpg 1024w\" sizes=\"auto, (max-width: 683px) 100vw, 683px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 12%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Jak zosta\u0142y ocenione detektory obrazu<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Model obrazu<\/th>\n <th>Kategoria<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>ChatGPT <span style=\"font-size:0.75rem; color:#888; display:block;\">Pe\u0142na generacja<\/span><\/td>\n <td>Og\u00f3lne<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">77.99% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">0.98% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">1.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p style=\"font-size:17px\">Dla mnie (a codziennie widz\u0119 wiele tre\u015bci generowanych przez sztuczn\u0105 inteligencj\u0119) obraz wygl\u0105da ca\u0142kiem przekonuj\u0105co na pierwszy rzut oka. Gdyby\u015b klikn\u0105\u0142 na czyj\u015b profil w mediach spo\u0142eczno\u015bciowych, przewin\u0105\u0142 jego kana\u0142 i zobaczy\u0142 ten obraz, czy od razu wyr\u00f3\u017cnia\u0142by si\u0119 jako fa\u0142szywy wygenerowany przez sztuczn\u0105 inteligencj\u0119? Aby by\u0107 uczciwym, stara\u0142em si\u0119 by\u0107 kreatywny z podpowiedzi\u0105. Oto ten, kt\u00f3rego u\u017cy\u0142em do wygenerowania tego obrazu:<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p class=\"has-text-align-center has-vivid-cyan-blue-color has-text-color has-link-color has-medium-font-size wp-elements-99b76fc78211de13a760aff177dd504c\">“<em><strong>Wygeneruj mi zdj\u0119cie w stylu migawki aparatu fotograficznego przedstawiaj\u0105ce m\u0119\u017cczyzn\u0119 stoj\u0105cego na szczycie dachu i patrz\u0105cego w d\u00f3\u0142 na ulic\u0119, noc, jest o\u015bwietlony lamp\u0105 b\u0142yskow\u0105 aparatu, s\u0105 \u015blady \u015bwiat\u0142a z samochod\u00f3w poni\u017cej, kobieta stoi za nim z r\u0119k\u0105 na ustach, u\u015bmiechni\u0119ta, estetyka zdj\u0119cia przypomina zdj\u0119cie zrobione migawk\u0105 aparatu w 2009 roku przez przyjaci\u00f3\u0142 z college'u.<\/strong><\/em>“<\/p>\n<\/blockquote>\n<\/blockquote>\n<\/blockquote>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p style=\"font-size:17px\">Niezale\u017cnie od tego, jak bardzo by\u0142em kreatywny lub ile szczeg\u00f3\u0142\u00f3w zawar\u0142em w mojej podpowiedzi, TruthScan i Sight Engine nadal poprawnie klasyfikowa\u0142y dane wyj\u015bciowe jako wygenerowane przez AI. Detektor AI or Not by\u0142 blisko, ale wci\u0105\u017c nie trafia\u0142. <\/p>\n\n\n\n<p style=\"font-size:17px\">Winston i WasItAI zdecydowanie si\u0119 mylili, klasyfikuj\u0105c obraz ChatGPT jako prawdziwy.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#2. \"Fa\u0142szywy paragon\" (wygenerowany przez ChatGPT)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"683\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest-683x1024.jpg\" alt=\"Obraz przedstawiaj\u0105cy co\u015b, co wygl\u0105da na poplamiony paragon\" class=\"wp-image-5291\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest-683x1024.jpg 683w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest-200x300.jpg 200w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest-768x1152.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest-8x12.jpg 8w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest.jpg 1024w\" sizes=\"auto, (max-width: 683px) 100vw, 683px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 12%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Jak zosta\u0142y ocenione detektory obrazu<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Model obrazu<\/th>\n <th>Kategoria<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>ChatGPT <span style=\"font-size:0.75rem; color:#888; display:block;\">Pe\u0142na generacja<\/span><\/td>\n <td>Oszustwo<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">19.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">94.48% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">0.04% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">1.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Chcia\u0142em wygenerowa\u0107 fa\u0142szywy paragon za pomoc\u0105 ChatGPT, aby zobaczy\u0107, jak realistycznie b\u0119dzie wygl\u0105da\u0142. Mo\u017cna zauwa\u017cy\u0107, \u017ce na obrazie brakuje prawdziwego adresu, ale tekstura nadal wygl\u0105da nieco realistycznie. Powiedzia\u0142bym, \u017ce ten obraz jest mniej przekonuj\u0105cy ni\u017c pierwszy, kt\u00f3ry wygenerowa\u0142em za pomoc\u0105 ChatGPT, ale wi\u0119kszo\u015b\u0107 detektor\u00f3w nadal go analizowa\u0142a.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li style=\"font-size:17px\">Ze wszystkich kontroli, kt\u00f3re przeprowadzi\u0142em na tym obrazie, <strong>TruthScan<\/strong> by\u0142 najdok\u0142adniejszy. <\/li>\n\n\n\n<li style=\"font-size:17px\"><strong>Silnik celowniczy<\/strong> Tutaj naprawd\u0119 zawiod\u0142a, pokazuj\u0105c s\u0142ab\u0105 ocen\u0119 obrazu AI zwi\u0105zanego z dokumentami.<\/li>\n\n\n\n<li style=\"font-size:17px\"><strong>Sztuczna inteligencja czy nie<\/strong> wypad\u0142a zauwa\u017calnie lepiej w tej drugiej analizie.<\/li>\n\n\n\n<li style=\"font-size:17px\"><strong>Winston AI <\/strong>wypad\u0142 najgorzej, a WasItAI r\u00f3wnie\u017c ca\u0142kowicie zawi\u00f3d\u0142.<\/li>\n<\/ul>\n\n\n\n<p style=\"font-size:17px\">Bior\u0105c pod uwag\u0119, \u017ce ten obraz jest powi\u0105zany z kategori\u0105 oszustwa, niepokoj\u0105ce jest to, \u017ce wi\u0119kszo\u015b\u0107 detektor\u00f3w b\u0142\u0119dnie go sklasyfikowa\u0142a. <\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Monit, kt\u00f3rego u\u017cy\u0142em do wygenerowania obrazu<\/h3>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-vivid-cyan-blue-color has-text-color has-link-color wp-elements-b58fc16e4a944c203389dba029f958d9\"><strong>“<em>Wygeneruj obraz paragonu z Best Buy, kt\u00f3ry m\u00f3wi, \u017ce \u0142\u0105czna kwota wydanych pieni\u0119dzy wynosi miliard dolar\u00f3w i jest na nim plama po kawie.<\/em>“<\/strong><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p style=\"font-size:17px\">Najwyra\u017aniej napisa\u0142em szybki i prosty monit. Mimo to jako\u015b\u0107 danych wyj\u015bciowych ChatGPT jest atrakcyjna wizualnie go\u0142ym okiem. Wydaje si\u0119, \u017ce nie jest tak atrakcyjna dla dw\u00f3ch z pi\u0119ciu testowanych przeze mnie detektor\u00f3w.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#3: \"Pakiet zagro\u017ce\u0144\" (wygenerowany przy u\u017cyciu Nano Banana)<\/h2>\n\n\n\n<p><strong><em>Zamiast kaza\u0107 Nano Banana wygenerowa\u0107 oryginalny obraz, da\u0142em mu jeden do zmiany. <\/em><\/strong><\/p>\n\n\n\n<p style=\"font-size:17px\">Pomys\u0142 polega\u0142 na zademonstrowaniu, jak kto\u015b mo\u017ce \u0142atwo wykorzysta\u0107 sztuczn\u0105 inteligencj\u0119 do stworzenia fa\u0142szywych dowod\u00f3w i twierdzenia, \u017ce otrzyma\u0142 uszkodzon\u0105\/niebezpieczn\u0105 paczk\u0119. Najpierw zrobi\u0142em zdj\u0119cie iPhone'em 15 pustej paczki Amazon, kt\u00f3r\u0105 znalaz\u0142em.<\/p>\n\n\n\n<p><strong>Oto zdj\u0119cie, kt\u00f3re zrobi\u0142em:<\/strong><\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"759\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package-759x1024.jpg\" alt=\"Zdj\u0119cie paczki Amazon\" class=\"wp-image-5293\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package-759x1024.jpg 759w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package-222x300.jpg 222w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package-768x1036.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package-9x12.jpg 9w\" sizes=\"auto, (max-width: 759px) 100vw, 759px\" \/><\/figure>\n\n\n\n<p style=\"font-size:17px\">Nast\u0119pnie przes\u0142a\u0142em zrobione przeze mnie zdj\u0119cie do Nano Banana i wyda\u0142em nast\u0119puj\u0105ce polecenie:<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-vivid-cyan-blue-color has-text-color has-link-color wp-elements-41ee9a66efaa63aab6aedf0d2e80cce1\"><strong>“<em>Chc\u0119, aby\u015b edytowa\u0142 to zdj\u0119cie, nie zmieniaj niczego opr\u00f3cz miejsca, w kt\u00f3rym znajduje si\u0119 etykieta, dodaj do niej uszkodzenia i dodaj czarne plamy szlamu na pude\u0142ku, kt\u00f3re wygl\u0105daj\u0105 jak prawdziwe plamy smaru.<\/em>“<\/strong><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\">Wydajno\u015b\u0107 Nano Bana:<\/h3>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"747\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged-747x1024.jpg\" alt=\"Zdj\u0119cie przedstawiaj\u0105ce paczk\u0119\" class=\"wp-image-5292\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged-747x1024.jpg 747w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged-219x300.jpg 219w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged-768x1052.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged-9x12.jpg 9w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged.jpg 864w\" sizes=\"auto, (max-width: 747px) 100vw, 747px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 12%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Jak zosta\u0142y ocenione detektory obrazu<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Model obrazu<\/th>\n <th>Kategoria<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana <span style=\"font-size:0.75rem; color:#888; display:block;\">Edycja Deepfake<\/span><\/td>\n <td>Oszustwo<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">98.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.10% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">31.98% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Olejopodobny szlam generowany przez Nano Banana na wyj\u015bciu mia\u0142 przekonuj\u0105cy brudny po\u0142ysk, wraz z efektem przemoczenia - wcale nie kresk\u00f3wkowy. Na szcz\u0119\u015bcie wi\u0119kszo\u015b\u0107 wykrywaczy, przez kt\u00f3re przeszed\u0142em, poprawnie oznaczy\u0142a obraz jako fa\u0142szywy. Jedynym wyj\u0105tkiem by\u0142 Winston, kt\u00f3ry ponownie wypad\u0142 s\u0142abo, b\u0142\u0119dnie klasyfikuj\u0105c obraz jako autentyczny. AI or Not, TruthScan i WasItAI by\u0142y najdok\u0142adniejszymi detektorami, kt\u00f3re wykry\u0142y ten obraz.<\/p>\n\n\n\n<p style=\"font-size:17px\">Niepokoj\u0105ce jest to, jak proste wydaje si\u0119 przes\u0142anie obrazu do chatbota i jego ca\u0142kowita zmiana. Mog\u0119 sobie wyobrazi\u0107 oszust\u00f3w i scammer\u00f3w korzystaj\u0105cych z narz\u0119dzi do generowania obraz\u00f3w, aby pr\u00f3bowa\u0107 sk\u0142ada\u0107 fa\u0142szywe roszczenia o zwrot pieni\u0119dzy lub anga\u017cowa\u0107 si\u0119 w oszustwa na rynku.<\/p>\n\n\n\n<p style=\"font-size:17px\">Je\u015bli platformy handlu elektronicznego wymagaj\u0105 jedynie obrazu jako dowodu do zainicjowania roszczenia o uszkodzon\u0105 paczk\u0119, to bez \u017cadnego niezawodnego sposobu wykrywania manipulacji AI s\u0105 one nara\u017cone na powa\u017cny wektor ataku. Na przyk\u0142ad oficjalna strona Amazon <a href=\"https:\/\/sellercentral.amazon.com\/help\/hub\/reference\/external\/GYW2EV9FEMJ3JJGR?locale=en-US\" target=\"_blank\" rel=\"noopener\">polityka zwrot\u00f3w przedmiot\u00f3w<\/a> dos\u0142ownie stwierdza<strong>, “<em>Materia\u0142y niebezpieczne, w tym \u0142atwopalne ciecze lub gazy, nie podlegaj\u0105 zwrotowi.<\/em>.”<\/strong><\/p>\n\n\n\n<p style=\"font-size:17px\">Je\u015bli wi\u0119c faktycznie otrzyma\u0142e\u015b paczk\u0119 pokryt\u0105 olejem lub t\u0142ustym czarnym szlamem, nie musia\u0142by\u015b jej zwraca\u0107, ale nadal m\u00f3g\u0142by\u015b otrzyma\u0107 zwrot pieni\u0119dzy. Je\u015bli z jakiego\u015b powodu faktycznie by ci si\u0119 to przydarzy\u0142o, Amazon prawdopodobnie poprosi\u0142by o dow\u00f3d w postaci zdj\u0119cia. Czy widzisz problem, jaki mog\u0105 stanowi\u0107 obrazy generowane przez sztuczn\u0105 inteligencj\u0119, gdy s\u0105 one nie do odr\u00f3\u017cnienia od prawdziwych?<\/p>\n\n\n\n<h2 class=\"wp-block-heading has-text-align-left\">#4. \"Cockroach Meal\" (wygenerowane przy u\u017cyciu Nano Banana)<\/h2>\n\n\n\n<p style=\"font-size:17px\">Ten kolejny obraz AI mo\u017ce sprawi\u0107, \u017ce stracisz apetyt. Jak si\u0119 okazuje, istniej\u0105 udokumentowane przypadki os\u00f3b u\u017cywaj\u0105cych sztucznej inteligencji do otrzymywania <a href=\"https:\/\/dailydot.com\/ai-generated-undercooked-food-doordash\" data-type=\"link\" data-id=\"https:\/\/dailydot.com\/ai-generated-undercooked-food-doordash\" target=\"_blank\" rel=\"noopener\">nieuczciwe zwroty koszt\u00f3w<\/a> w aplikacjach z dostaw\u0105 jedzenia. W czwartym te\u015bcie wykona\u0142em te same kroki, co w te\u015bcie trzecim. Tym razem zrobi\u0142em zdj\u0119cie mojego pustego pude\u0142ka po obiedzie.<\/p>\n\n\n\n<p><strong>Oto prawdziwy obraz:<\/strong><\/p>\n\n\n\n<figure class=\"wp-block-image size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"768\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/IMG_6145-768x1024.jpg\" alt=\"Zdj\u0119cie pustego pude\u0142ka po jedzeniu na wynos.\" class=\"wp-image-5296\" style=\"width:462px;height:auto\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/IMG_6145-768x1024.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/IMG_6145-225x300.jpg 225w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/IMG_6145-scaled.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/IMG_6145-9x12.jpg 9w\" sizes=\"auto, (max-width: 768px) 100vw, 768px\" \/><\/figure>\n\n\n\n<p style=\"font-size:17px\">Nast\u0119pnie umie\u015bci\u0142em prawdziwe zdj\u0119cie w Nano Banana i wyda\u0142em nast\u0119puj\u0105ce polecenie:<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-vivid-cyan-blue-color has-text-color has-link-color wp-elements-e9f00ca4e893e7742bce463148a639f5\"><strong><em>\"Edytuj ten obraz. Nie zmieniaj niczego poza dodaniem makaronu w miejscu, gdzie znajduje si\u0119 folia i dodaniem kilku ma\u0142ych karaluch\u00f3w w jedzeniu. Niech wszystko wygl\u0105da realistycznie, a nie kresk\u00f3wkowo\". <\/em><\/strong><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\">Wyj\u015bcie Nano Banana:<\/h3>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"747\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Cockroach-meal.jpg\" alt=\"Zmienione przez sztuczn\u0105 inteligencj\u0119 zdj\u0119cie pude\u0142ka z jedzeniem na wynos z makaronem i robakami w \u015brodku\" class=\"wp-image-5295\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Cockroach-meal.jpg 747w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Cockroach-meal-219x300.jpg 219w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Cockroach-meal-9x12.jpg 9w\" sizes=\"auto, (max-width: 747px) 100vw, 747px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 12%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Jak zosta\u0142y ocenione detektory obrazu<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Model obrazu<\/th>\n <th>Kategoria<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana <span style=\"font-size:0.75rem; color:#888; display:block;\">Edycja Deepfake<\/span><\/td>\n <td>Oszustwo<\/td>\n <td><span class=\"score-pill score-pass\">97.15% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">18.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">85.87% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">72.29% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">1.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Po raz kolejny TruthScan pewnie wykry\u0142 media wygenerowane przez AI. Sight Engine i WasitAI ca\u0142kowicie zawiod\u0142y. AI or Not by\u0142 drugim najdok\u0142adniejszym wykrywaczem (z pewno\u015bci\u0105 85% AI) podczas skanowania tego obrazu AI, a Winston wypad\u0142 lepiej ni\u017c w trzecim te\u015bcie (z pewno\u015bci\u0105 72% AI).<\/p>\n\n\n\n<p style=\"font-size:17px\">Na tym obrazie makaron i karaluchy zosta\u0142y w ca\u0142o\u015bci wygenerowane przez sztuczn\u0105 inteligencj\u0119; jednak dwa detektory nie wydawa\u0142y si\u0119 tak my\u015ble\u0107, a tylko jeden (TruthScan) by\u0142 powy\u017cej 90% pewny manipulacji przez sztuczn\u0105 inteligencj\u0119.<\/p>\n\n\n\n<p style=\"font-size:17px\">Niepokoj\u0105ce by\u0142o to, jak proste by\u0142o stworzenie tego fa\u0142szywego obrazu. Od przes\u0142ania zdj\u0119cia, kt\u00f3re zrobi\u0142em, do podpowiedzi Nano Banana i otrzymania wyniku, w sumie zaj\u0119\u0142o to tylko oko\u0142o 2 minut. Koszmar dla aplikacji dostarczaj\u0105cych jedzenie; marzenie dla g\u0142odnych k\u0142amc\u00f3w. Czynnik oszustwa.<\/p>\n\n\n\n<p class=\"has-vivid-red-color has-text-color has-link-color wp-elements-4c89b6ee3afcbdd6b6617c150e5b2e03\"><strong>UJAWNIENIE: Niekt\u00f3re z obraz\u00f3w generowanych przez sztuczn\u0105 inteligencj\u0119 zawartych w poni\u017cszych testach mia\u0142y na celu zademonstrowanie, w jaki spos\u00f3b obrazy AI mog\u0105 by\u0107 wykorzystywane do dezinformacji. Zawieraj\u0105 one kontrowersyjne obrazy i wra\u017cliwe tematy<\/strong>, <strong>i NIE maj\u0105 na celu reprezentowania jakichkolwiek konkretnych<\/strong> <strong>pogl\u0105dy lub ideologie.<\/strong><\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><strong>#5<\/strong>. \"Autowandalizm\" (wygenerowane przez Nano Banana)<\/h2>\n\n\n\n<p style=\"font-size:17px\">Znalaz\u0142em prawdziwe zdj\u0119cie samochodu na sprzeda\u017c online. Ca\u0142kiem przyziemne. Ale co by by\u0142o, gdyby kto\u015b chcia\u0142 wykorzysta\u0107 ten obraz do rozpowszechniania dezinformacji lub z\u0142o\u017cenia fa\u0142szywego roszczenia ubezpieczeniowego? Niestety, generatywna sztuczna inteligencja sprawia, \u017ce jest to szybkie i \u0142atwe.<\/p>\n\n\n\n<p style=\"font-size:17px\"><strong>W pi\u0105tym te\u015bcie wzi\u0105\u0142em prawdziwy obraz samochodu i da\u0142em Nano Banana nast\u0119puj\u0105ce polecenie:<\/strong><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-vivid-cyan-blue-color has-text-color has-link-color wp-elements-d06074fe9533ce61c08bcf9750cd398d\">“<strong><em>Nie zmieniaj niczego w samochodzie, z wyj\u0105tkiem tego, \u017ce chcesz doda\u0107 do niego farb\u0119 w sprayu z napisem \"Wolna Palestyna\", a przednia szyba jest p\u0119kni\u0119ta. Reflektory s\u0105 rozbite, lusterko jest z\u0142amane i zwisa w d\u00f3\u0142, a dwie szyby z boku samochodu s\u0105 uszkodzone.<\/em><\/strong> <strong>rozbity<\/strong>.”<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\"><strong>Oto prawdziwy obraz:<\/strong><\/h3>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"940\" height=\"627\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/2000_honda_civic-si_2000_honda_civic-si_bbeed9f1-7b50-4314-a504-40be9411b8ab-mXhf1y-2.webp\" alt=\"Zdj\u0119cie niebieskiego samochodu\" class=\"wp-image-5298\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/2000_honda_civic-si_2000_honda_civic-si_bbeed9f1-7b50-4314-a504-40be9411b8ab-mXhf1y-2.webp 940w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/2000_honda_civic-si_2000_honda_civic-si_bbeed9f1-7b50-4314-a504-40be9411b8ab-mXhf1y-2-300x200.webp 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/2000_honda_civic-si_2000_honda_civic-si_bbeed9f1-7b50-4314-a504-40be9411b8ab-mXhf1y-2-768x512.webp 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/2000_honda_civic-si_2000_honda_civic-si_bbeed9f1-7b50-4314-a504-40be9411b8ab-mXhf1y-2-18x12.webp 18w\" sizes=\"auto, (max-width: 940px) 100vw, 940px\" \/><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">Obraz Nano Banana AI:<\/h3>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"683\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Protest-Car-Damage.jpg\" alt=\"Zmienione przez sztuczn\u0105 inteligencj\u0119 zdj\u0119cie niebieskiego samochodu, pokazuj\u0105ce p\u0119kni\u0119t\u0105 przedni\u0105 szyb\u0119 i uszkodzenia farby w sprayu z boku samochodu.\" class=\"wp-image-5299\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Protest-Car-Damage.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Protest-Car-Damage-300x200.jpg 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Protest-Car-Damage-768x512.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Protest-Car-Damage-18x12.jpg 18w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Jak zosta\u0142y ocenione detektory obrazu<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Model obrazu<\/th>\n <th>Kategoria<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana <span style=\"font-size:0.75rem; color:#888; display:block;\">Edycja Deepfake<\/span><\/td>\n <td style=\"font-size:0.82rem;\">Oszustwo \/ Dezinformacja<\/td>\n <td><span class=\"score-pill score-pass\">97.48% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">18.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">89.52% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">0.02% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Wyniki wykrywania dla tego wyj\u015bcia by\u0142y mieszane. W tym te\u015bcie WasItAI (zaskakuj\u0105co) uzyska\u0142 najwy\u017cszy wynik wykrywania AI (99%). TruthScan mia\u0142 drugi najwy\u017cszy wynik (97%). Sight Engine znacz\u0105co nie wykry\u0142 \u017cadnych syntetycznych element\u00f3w na obrazie, a Winston poradzi\u0142 sobie najgorzej ze wszystkich detektor\u00f3w (przypisuj\u0105c mniej ni\u017c 1% szansy na zaanga\u017cowanie). AI or Not poradzi\u0142o sobie dobrze, ale dok\u0142adno\u015b\u0107 nadal by\u0142a poni\u017cej 90%.<br><br>Porozmawiajmy o obrazie. Dwa najbardziej niepokoj\u0105ce aspekty tre\u015bci stworzonych przez sztuczn\u0105 inteligencj\u0119 lub zmienionych w ten spos\u00f3b to fakt, \u017ce oba wi\u0105\u017c\u0105 si\u0119 z oszustwem. Po pierwsze, wykorzystanie sztucznej inteligencji do zmiany przyziemnych obraz\u00f3w w celu stworzenia tre\u015bci o charakterze politycznym jest szybkie, \u0142atwe i wygl\u0105da realistycznie. Niezale\u017cnie od tego, czy osoby tworz\u0105ce tego typu tre\u015bci maj\u0105 na celu jedynie zaanga\u017cowanie, czy te\u017c s\u0105 politycznymi prowokatorami, problem polega na tym, \u017ce s\u0105 to fa\u0142szywe informacje przedstawiane jako autentyczne.<\/p>\n\n\n\n<p style=\"font-size:17px\">Drugim sposobem wykorzystania takich narz\u0119dzi do generowania obraz\u00f3w jest oszustwo. Je\u015bli oszust chce z\u0142o\u017cy\u0107 fa\u0142szywe roszczenie z tytu\u0142u ubezpieczenia autocasco, b\u0119dzie musia\u0142 przedstawi\u0107 fa\u0142szywe dowody. Zamiast sp\u0119dza\u0107 godziny w Photoshopie na edytowaniu obraz\u00f3w i poprawianiu dowod\u00f3w, mog\u0105 u\u017cy\u0107 generator\u00f3w obraz\u00f3w, aby szybko stworzy\u0107 fa\u0142szywy dow\u00f3d.<\/p>\n\n\n\n<p style=\"font-size:17px\">W powy\u017cszym przyk\u0142adzie obrazu AI zauwa\u017c, jak oszust mo\u017ce wykorzysta\u0107 podobny proces do z\u0142o\u017cenia fa\u0142szywego roszczenia zgodnie z Progressive's <a href=\"https:\/\/www.progressive.com\/answers\/does-car-insurance-cover-vandalism\/\" data-type=\"link\" data-id=\"https:\/\/www.progressive.com\/answers\/does-car-insurance-cover-vandalism\/\" target=\"_blank\" rel=\"noopener\">Kompleksowa polityka ochrony<\/a>kt\u00f3ry w szczeg\u00f3lno\u015bci obejmuje:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Ci\u0119cie lub <a href=\"https:\/\/www.progressive.com\/answers\/does-car-insurance-cover-tire-damage\/\" target=\"_blank\" rel=\"noopener\">uszkodzone opony<\/a><\/strong><\/li>\n\n\n\n<li><strong>Uszkodzone szyby, reflektory lub tylne \u015bwiat\u0142a<\/strong><\/li>\n\n\n\n<li><strong>Uszkodzenia farby w sprayu<\/strong><\/li>\n\n\n\n<li><strong><a href=\"https:\/\/www.progressive.com\/answers\/does-car-insurance-cover-scratches\/\" target=\"_blank\" rel=\"noopener\">Wgniecenia lub zadrapania<\/a> przed w\u0142amaniem do samochodu<\/strong><\/li>\n\n\n\n<li><strong>Wsypywanie cukru lub innych substancji do zbiornika paliwa<\/strong><\/li>\n<\/ul>\n\n\n\n<p>Oczywi\u015bcie sk\u0142adanie fa\u0142szywych dowod\u00f3w jest nie tylko zwodnicze, ale tak\u017ce ca\u0142kowicie nielegalne. W idealnym \u015bwiecie nikt nie \u0142ama\u0142by prawa; w rzeczywisto\u015bci przest\u0119pstwa zdarzaj\u0105 si\u0119 ka\u017cdego dnia. A bez niezawodnego wykrywania, firmy ubezpieczeniowe ryzykuj\u0105, \u017ce zostan\u0105 oszukane na miliony, poniewa\u017c ro\u015bnie liczba fa\u0142szywych roszcze\u0144 ubezpieczeniowych z wykorzystaniem sztucznej inteligencji.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#6. \"Soda Musk i Don\" (wygenerowane przy u\u017cyciu Nano Banana)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"559\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-1024x559.jpeg\" alt=\"Wygenerowany przez sztuczn\u0105 inteligencj\u0119 obraz Elona Muska i Donalda Trumpa\" class=\"wp-image-5300\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-1024x559.jpeg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-300x164.jpeg 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-768x419.jpeg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-1536x838.jpeg 1536w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-2048x1117.jpeg 1877w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-18x10.jpeg 18w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-scaled.jpeg 1878w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Jak zosta\u0142y ocenione detektory obrazu<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Model obrazu<\/th>\n <th>Kategoria<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana <span style=\"font-size:0.75rem; color:#888; display:block;\">Pe\u0142na generacja<\/span><\/td>\n <td style=\"font-size:0.82rem;\">Disinfo<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">0.24% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">86.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Po raz kolejny u\u017cy\u0142em Nano Banana do tego testu, a celem by\u0142o zademonstrowanie fa\u0142szywych, politycznie na\u0142adowanych obraz\u00f3w o g\u0142upszym wyd\u017awi\u0119ku. TruthScan, Sight Engine i AIorNot oznaczy\u0142y wygenerowany obraz z ocen\u0105 wykrywania AI 99%. WasitAI nie by\u0142 tak pewny siebie - wykrywaj\u0105c obraz wygenerowany przez sztuczn\u0105 inteligencj\u0119 z pewno\u015bci\u0105 tylko 86%. Winston ca\u0142kowicie zawi\u00f3d\u0142, przypisuj\u0105c wynik AI na poziomie 0,2%. Aby wygenerowa\u0107 obraz, da\u0142em Nano Banana ten monit:<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-vivid-cyan-blue-color has-text-color has-link-color wp-elements-4f7e002ecaefec24bbebdca3101fff42\"><strong>“<em>Wygeneruj zdj\u0119cie, na kt\u00f3rym Elon Musk rozlewa du\u017cy kubek napoju gazowanego jak Big Gulp na swoj\u0105 koszul\u0119 i wariuje, podczas gdy Donald Trump siedzi obok niego i si\u0119 \u015bmieje. S\u0105 w samolocie odrzutowym. To jest ta scena. A zdj\u0119cie powinno wygl\u0105da\u0107 realistycznie, jakby zosta\u0142o zrobione aparatem fotograficznym i zawiera\u0107 artefakty, takie jak lampa b\u0142yskowa aparatu odbijaj\u0105ca si\u0119 w oknie samolotu i na zewn\u0105trz samolotu. Jest noc. Powinno wygl\u0105da\u0107 hiperrealistycznie.<\/em>“<\/strong><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p style=\"font-size:17px\">Ilekro\u0107 jeden z tych program\u00f3w nie wykryje sztucznej inteligencji, oznaczaj\u0105 go jako autentyczny. Jakkolwiek g\u0142upi mo\u017ce by\u0107 ten przyk\u0142ad obrazu z testu, nie ma nic zabawnego w my\u015bleniu, \u017ce deepfake jest prawdziwy.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#7. \"Still Alive\" (wygenerowane przez Nano Banana)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"765\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3Christian_Ransom_Photo.jpg\" alt=\"Wygenerowany przez sztuczn\u0105 inteligencj\u0119 obraz m\u0119\u017cczyzny trzymaj\u0105cego znak z napisem "pomoc"\" class=\"wp-image-5301\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3Christian_Ransom_Photo.jpg 765w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3Christian_Ransom_Photo-224x300.jpg 224w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3Christian_Ransom_Photo-9x12.jpg 9w\" sizes=\"auto, (max-width: 765px) 100vw, 765px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Jak zosta\u0142y ocenione detektory obrazu<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Model obrazu<\/th>\n <th>Kategoria<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana <span style=\"font-size:0.75rem; color:#888; display:block;\">Pe\u0142na generacja<\/span><\/td>\n <td style=\"font-size:0.82rem;\">Deepfake<\/td>\n <td><span class=\"score-pill score-pass\">97.49% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">95.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.27% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">0.15% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">1.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Czy uwa\u017casz, \u017ce ten obraz jest niepokoj\u0105cy? Aby wygenerowa\u0107 ten deepfake, u\u017cy\u0142em Nano Banana i da\u0142em mu swoje zdj\u0119cie. Nast\u0119pnie da\u0142em mu d\u0142ug\u0105 podpowied\u017a, kt\u00f3ra mia\u0142a postawi\u0107 mnie w sytuacji sugeruj\u0105cej, \u017ce zosta\u0142em uprowadzony. Mo\u017ce dlatego, \u017ce to zdj\u0119cie pokazuje <em>ja<\/em> w kompromituj\u0105cej pozycji. Co by by\u0142o, gdyby kto\u015b stworzy\u0142 co\u015b takiego, wys\u0142a\u0142 to mojej matce i zagrozi\u0142, \u017ce mnie skrzywdzi, je\u015bli nie wy\u015ble mu pieni\u0119dzy? Straszne rzeczy. <\/p>\n\n\n\n<p class=\"has-small-font-size\"><em>(Pe\u0142ny monit, kt\u00f3rego u\u017cy\u0142em w tym celu, mo\u017cna znale\u017a\u0107 w pe\u0142nym zestawie danych testowych na ko\u0144cu tego artyku\u0142u).<\/em><\/p>\n\n\n\n<p style=\"font-size:17px\"><br>Na szcz\u0119\u015bcie ten obraz ma bardzo realistyczny wygl\u0105d, <strong>Trzy na pi\u0119\u0107 detektor\u00f3w zidentyfikowa\u0142o je jako wygenerowane przez sztuczn\u0105 inteligencj\u0119.<\/strong> <\/p>\n\n\n\n<p style=\"font-size:17px\">AI or Not uzyska\u0142o wynik 99%, TruthScan 97%, a Sight Engine 95%. Niestety, Winston i WasItAI sklasyfikowa\u0142y obraz jako prawdziwy.<\/p>\n\n\n\n<p style=\"font-size:17px\">Powinienem wyja\u015bni\u0107, \u017ce testy te nie zosta\u0142y przeprowadzone na moim osobistym koncie Gemini. Nano Banana nigdy nie poprosi\u0142a mnie o udowodnienie, \u017ce jestem osob\u0105 na zdj\u0119ciu. Ka\u017cdy m\u00f3g\u0142 pobra\u0107 moje zdj\u0119cie z Internetu, umie\u015bci\u0107 je w Nano Banana i zleci\u0107 stworzenie tego typu obrazu. <\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#8. \"Abuse of a Ben\" (wygenerowane przy u\u017cyciu Nano Banana)<\/h2>\n\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"200\" height=\"200\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/c945d528-730f-48c1-bdca-30e5799fc1aa_thumb-1.jpg\" alt=\"\" class=\"wp-image-5307\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/c945d528-730f-48c1-bdca-30e5799fc1aa_thumb-1.jpg 200w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/c945d528-730f-48c1-bdca-30e5799fc1aa_thumb-1-150x150.jpg 150w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/c945d528-730f-48c1-bdca-30e5799fc1aa_thumb-1-12x12.jpg 12w\" sizes=\"auto, (max-width: 200px) 100vw, 200px\" \/><\/figure>\n<\/div>\n\n\n<p style=\"font-size:17px\">Poznaj Bena. Wi\u0119kszo\u015b\u0107 czasu sp\u0119dza na zarz\u0105dzaniu r\u00f3\u017cnymi dzia\u0142ami w firmach, w kt\u00f3rych pracuje. Powy\u017csze zdj\u0119cie to (prawdziwy) elegancki firmowy headshot Bena. Je\u015bli trudno si\u0119 domy\u015bli\u0107, Ben jest profesjonalist\u0105. Bez wzgl\u0119du na to, jakie ma pogl\u0105dy polityczne, zachowuje je dla siebie, a jego rozmowy w pracy dotycz\u0105 tylko pracy. Jednak w ka\u017cdej chwili podobizna Bena mo\u017ce zosta\u0107 wykorzystana do rozpowszechniania propagandy politycznej lub wiadomo\u015bci bez jego zgody.<\/p>\n\n\n\n<p style=\"font-size:17px\">Ostatnio pojawi\u0142o si\u0119 wiele doniesie\u0144 o tym, jak generowane przez sztuczn\u0105 inteligencj\u0119 filmy z udzia\u0142em agent\u00f3w ICE i protest\u00f3w rozprzestrzeniaj\u0105 si\u0119 w Internecie. Na podstawie wszystkich dotychczasowych test\u00f3w doszed\u0142em do wniosku, \u017ce tworzenie fa\u0142szywej propagandy politycznej za pomoc\u0105 sztucznej inteligencji jest naprawd\u0119 \u0142atwe. Postanowi\u0142em przetestowa\u0107 ponownie, tym razem u\u017cywaj\u0105c Bena. <\/p>\n\n\n\n<p style=\"font-size:17px\">Za zgod\u0105 Bena wykorzysta\u0142em jego zdj\u0119cie g\u0142owy do stworzenia deepfake'a. <strong>Oto obraz wygenerowany przez sztuczn\u0105 inteligencj\u0119, kt\u00f3ry stworzy\u0142 Nano Banana:<\/strong><\/p>\n\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-full is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x.jpg\" alt=\"Zmienione przez sztuczn\u0105 inteligencj\u0119 zdj\u0119cie m\u0119\u017cczyzny o imieniu Ben, trzymaj\u0105cego znak z napisem Abolish ICE.\" class=\"wp-image-5308\" style=\"width:537px;height:auto\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x-300x300.jpg 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x-150x150.jpg 150w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x-768x768.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x-12x12.jpg 12w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n<\/div>\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Jak zosta\u0142y ocenione detektory obrazu<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Model obrazu<\/th>\n <th>Kategoria<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana<\/td>\n <td style=\"font-size:0.78rem;\">Deepfake \/ Disinfo<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">97.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">98.88% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">5.74% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Zaj\u0119\u0142o to trzydzie\u015bci sekund. Wszystko, co musia\u0142em zrobi\u0107, to przeci\u0105gn\u0105\u0107 i upu\u015bci\u0107 jego zdj\u0119cie g\u0142owy do Gemini i poprosi\u0107 go o \"podniesienie znaku z napisem zniesienia ICE\". Wynik z Nano Banana nie jest tak fotorealistyczny jak inne, ale nadal wygl\u0105da wystarczaj\u0105co autentycznie, by by\u0107 przekonuj\u0105cym.<\/p>\n\n\n\n<p style=\"font-size:17px\"><span style=\"box-sizing: border-box; margin: 0px; padding: 0px;\">Na szcz\u0119\u015bcie cztery detektory dok\u0142adnie zidentyfikowa\u0142y obraz:\u00a0<strong>TruthScan<\/strong> (9<\/span>9% AI), Was It AI (99% AI), Sight Engine (97% AI) i AI or Not (97% AI). Winston AI zawi\u00f3d\u0142, przypisuj\u0105c tylko wynik 5% AI.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#9. \"Club Void\" (Wygenerowane przez Midjourney)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"771\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club-1024x771.jpg\" alt=\"U\u015bmiechni\u0119ta kobieta w bia\u0142ej sukience z kr\u00f3tkim r\u0119kawem stoi w s\u0142abo o\u015bwietlonym, zat\u0142oczonym pomieszczeniu. Zdj\u0119cie wydaje si\u0119 by\u0107 zrobione z lamp\u0105 b\u0142yskow\u0105, tworz\u0105c mglist\u0105, niebieskaw\u0105 atmosfer\u0119 ze s\u0142abymi cz\u0105steczkami widocznymi w powietrzu. Inne osoby w zwyk\u0142ych ubraniach stoj\u0105 w tle pod uszkodzonym lub przemys\u0142owo wygl\u0105daj\u0105cym sufitem.\" class=\"wp-image-5309\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club-1024x771.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club-300x226.jpg 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club-768x578.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club-16x12.jpg 16w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club.jpg 1232w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Jak zosta\u0142y ocenione detektory obrazu<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Model obrazu<\/th>\n <th>Kategoria<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>MidJourney <span style=\"font-size:0.75rem; color:#888; display:block;\">Pe\u0142na generacja<\/span><\/td>\n <td>Og\u00f3lne<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">89.21% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">1.56% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">70.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Wygenerowa\u0142em powy\u017cszy obraz za pomoc\u0105 Midjourney. Im d\u0142u\u017cej na niego patrzy\u0142em, tym bardziej wydawa\u0142 mi si\u0119 niepokoj\u0105cy. Estetyka jest mroczna i niesamowita. Wydaje si\u0119, \u017ce zza obiektu wy\u0142ania si\u0119 cienista obecno\u015b\u0107. Aby stworzy\u0107 ten obraz, u\u017cy\u0142em funkcji odniesienia estetycznego Midjourney i po prostu poprosi\u0142em, aby by\u0142a to \"dziewczyna stoj\u0105ca w klubie i u\u015bmiechaj\u0105ca si\u0119\". Pokaza\u0142em obraz kilku osobom i wszyscy my\u015bleli, \u017ce jest prawdziwy. <br><br>Wi\u0119kszo\u015b\u0107 wykrywaczy oznaczy\u0142a obraz jako wygenerowany przez sztuczn\u0105 inteligencj\u0119. TruthScan i Sight Engine oznaczy\u0142y go z wynikiem 99% AI, AI or Not powiedzia\u0142 89%. <\/p>\n\n\n\n<p style=\"font-size:17px\">WasItAI nie by\u0142 tak pewny jak detektory, klasyfikuj\u0105c obraz tylko z wynikiem 70% AI. <\/p>\n\n\n\n<p style=\"font-size:17px\">Detektor Winston AI ca\u0142kowicie zawi\u00f3d\u0142, daj\u0105c obrazowi wynik 1,56% AI. <\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#10 \"Ice Box\"<strong> <\/strong>(wygenerowane przez Midjourney)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"771\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE-1024x771.jpg\" alt=\"Rozmazane zdj\u0119cie przedstawiaj\u0105ce m\u0119\u017cczyzn\u0119 w poplamionej na czerwono koszuli, stoj\u0105cego przed kim\u015b.\" class=\"wp-image-5311\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE-1024x771.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE-300x226.jpg 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE-768x578.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE-16x12.jpg 16w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE.jpg 1232w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Jak zosta\u0142y ocenione detektory obrazu<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Model obrazu<\/th>\n <th>Kategoria<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>MidJourney <span style=\"font-size:0.75rem; color:#888; display:block;\">Pe\u0142na generacja<\/span><\/td>\n <td style=\"font-size:0.82rem;\">Disinfo<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">92.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">97.94% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">41.20% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">72.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Stworzy\u0142em ten obraz przy u\u017cyciu tej samej metody w po\u0142owie podr\u00f3\u017cy, co w poprzednim te\u015bcie. Tym razem celem by\u0142o wygenerowanie czego\u015b nieco ciemniejszego. TruthScan by\u0142 najdok\u0142adniejszym wykrywaczem, oznaczaj\u0105c obraz jako 99% AI. Sight Engine i AI or Not poprawnie wykry\u0142y obraz. Winston AI ca\u0142kowicie zawi\u00f3d\u0142, a WasItAI mia\u0142 znacznie ni\u017cszy wynik zaufania ni\u017c TruthScan, Sight Engine i AI or Not.<\/p>\n\n\n\n<p style=\"font-size:17px\">Powinienem powiedzie\u0107, \u017ce w tym przypadku wygenerowany przeze mnie obraz mo\u017cna wykorzysta\u0107 lub oprawi\u0107 na dwa sposoby. Z jednej strony, kto\u015b m\u00f3g\u0142by u\u017cy\u0107 tego (lub dowolnego innego o tematyce politycznej\/niepokoj\u00f3w spo\u0142ecznych) rodzaju medi\u00f3w w jakim\u015b projekcie artystycznym lub kreatywnym, aby z\u0142o\u017cy\u0107 o\u015bwiadczenie. W takim przypadku ryzyko wydaje si\u0119 mniejsze. G\u0142\u00f3wn\u0105 obaw\u0105, jak\u0105 mam w zwi\u0105zku z t\u0105 kategori\u0105 medi\u00f3w generowanych przez sztuczn\u0105 inteligencj\u0119, jest to, \u017ce s\u0105 one wykorzystywane przez nikczemne osoby, kt\u00f3re twierdz\u0105, \u017ce s\u0105 prawdziwe.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Podsumowanie i wyniki: Najlepsze wykrywacze obraz\u00f3w AI<\/h2>\n\n\n\n<p style=\"font-size:17px\">Ok, wi\u0119c teraz, gdy wy\u015bwietlili\u015bmy wszystkie testy, podsumuj\u0119: Wygenerowa\u0142em 10 obraz\u00f3w AI<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li style=\"font-size:17px\">U\u017cy\u0142em ChatGPT, Nano Banana i Midjourney do wygenerowania 10 obraz\u00f3w AI<\/li>\n\n\n\n<li style=\"font-size:17px\">Przetestowa\u0142em pi\u0119\u0107 wykrywaczy obraz\u00f3w AI, przepuszczaj\u0105c przez nie wszystkie wygenerowane przeze mnie obrazy AI.<\/li>\n\n\n\n<li style=\"font-size:17px\">TruthScan przeszed\u0142 wszystkie testy i by\u0142 najdok\u0142adniejszym wykrywaczem. AI or Not przeszed\u0142 8 z 10 test\u00f3w, wykazuj\u0105c pewn\u0105 niezawodno\u015b\u0107. Sight Engine nie przeszed\u0142 3 z 10 test\u00f3w i wykaza\u0142 si\u0119 og\u00f3lnie w\u0105tpliw\u0105 dok\u0142adno\u015bci\u0105. Was It AI nie zda\u0142 4 z 10 test\u00f3w i wykaza\u0142 si\u0119 s\u0142ab\u0105 dok\u0142adno\u015bci\u0105. Winston AI by\u0142 najmniej dok\u0142adnym wykrywaczem obraz\u00f3w AI, przechodz\u0105c tylko 3 z 10 test\u00f3w, konsekwentnie b\u0142\u0119dnie klasyfikuj\u0105c obrazy.<\/li>\n<\/ul>\n\n\n\n<!DOCTYPE html>\n<html lang=\"en\">\n<head>\n<meta charset=\"UTF-8\">\n<meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n<title>Detektor obraz\u00f3w AI - kompleksowe wyniki test\u00f3w<\/title>\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n *,*::before,*::after{box-sizing:border-box;margin:0;padding:0}\n\n body {\n background: #f4f3f0;\n min-height: 100vh;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n color: #2d2d3a;\n }\n\n \/* \u2500\u2500 wrapper \u2500\u2500 *\/\n .results-page {\n max-width: 860px;\n margin: 0 auto;\n padding: 3rem 1.25rem 4rem;\n }\n\n \/* \u2500\u2500 hero header \u2500\u2500 *\/\n .hero {\n text-align: center;\n margin-bottom: 2.75rem;\n }\n .hero-badge {\n display: inline-block;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.65rem;\n text-transform: uppercase;\n letter-spacing: 0.12em;\n color: #fff;\n background: #1a1a2e;\n padding: 6px 16px;\n border-radius: 100px;\n margin-bottom: 1.1rem;\n }\n .hero h1 {\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 2.4rem;\n color: #1a1a2e;\n letter-spacing: -0.03em;\n line-height: 1.15;\n margin-bottom: 0.6rem;\n }\n .hero p {\n font-size: 1.02rem;\n color: #6b6b7b;\n max-width: 560px;\n margin: 0 auto;\n line-height: 1.55;\n }\n\n \/* \u2500\u2500 accuracy cards row \u2500\u2500 *\/\n .accuracy-cards {\n display: grid;\n grid-template-columns: repeat(5, 1fr);\n gap: 10px;\n margin-bottom: 2.75rem;\n }\n .acc-card {\n background: #fff;\n border-radius: 14px;\n padding: 1.15rem 0.6rem 1rem;\n text-align: center;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.05),\n 0 6px 20px rgba(0,0,0,0.07),\n 0 0 0 1px rgba(0,0,0,0.03);\n position: relative;\n overflow: hidden;\n transition: transform 0.2s ease, box-shadow 0.2s ease;\n }\n .acc-card:hover {\n transform: translateY(-3px);\n box-shadow:\n 0 2px 6px rgba(0,0,0,0.06),\n 0 12px 32px rgba(0,0,0,0.1),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n .acc-card::before {\n content: '';\n position: absolute;\n top: 0; left: 0; right: 0;\n height: 4px;\n }\n .acc-card.rank-1::before { background: #16a34a; }\n .acc-card.rank-2::before { background: #a3e635; }\n .acc-card.rank-3::before { background: #facc15; }\n .acc-card.rank-4::before { background: #fb923c; }\n .acc-card.rank-5::before { background: #ef4444; }\n\n .acc-card .rank-num {\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 0.62rem;\n text-transform: uppercase;\n letter-spacing: 0.08em;\n color: #999;\n margin-bottom: 0.35rem;\n }\n .acc-card .tool-name {\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.92rem;\n color: #1a1a2e;\n margin-bottom: 0.65rem;\n line-height: 1.2;\n }\n .acc-ring {\n width: 72px; height: 72px;\n margin: 0 auto 0.55rem;\n position: relative;\n }\n .acc-ring svg { transform: rotate(-90deg); }\n .acc-ring .ring-bg { fill: none; stroke: #eee; stroke-width: 5; }\n .acc-ring .ring-fg { fill: none; stroke-width: 5; stroke-linecap: round; transition: stroke-dashoffset 1s ease; }\n .acc-ring .ring-label {\n position: absolute;\n inset: 0;\n display: flex;\n align-items: center;\n justify-content: center;\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 1.1rem;\n color: #1a1a2e;\n }\n .acc-card .acc-fraction {\n font-size: 0.78rem;\n color: #888;\n font-weight: 500;\n }\n\n \/* \u2500\u2500 section titles \u2500\u2500 *\/\n .section-title {\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 1.45rem;\n color: #1a1a2e;\n letter-spacing: -0.02em;\n margin-bottom: 1rem;\n }\n .section-subtitle {\n font-size: 0.88rem;\n color: #888;\n margin-bottom: 1.25rem;\n line-height: 1.5;\n }\n\n \/* \u2500\u2500 master table \u2500\u2500 *\/\n .master-table-wrap {\n border-radius: 14px;\n overflow: hidden;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n margin-bottom: 2.75rem;\n background: #fff;\n }\n .master-table {\n width: 100%;\n border-collapse: collapse;\n table-layout: fixed;\n }\n .master-table thead th {\n background: #1a1a2e;\n color: #fff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.62rem;\n text-transform: uppercase;\n letter-spacing: 0.05em;\n padding: 13px 5px;\n text-align: center;\n border: none;\n position: sticky;\n top: 0;\n z-index: 2;\n }\n .master-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 14%;\n }\n .master-table thead th:nth-child(2) { width: 13%; }\n .master-table thead th:nth-child(3) { width: 9%; }\n\n .master-table tbody tr {\n transition: background 0.18s ease;\n }\n .master-table tbody tr:hover {\n background: rgba(26,26,46,0.025);\n }\n .master-table tbody td {\n padding: 11px 5px;\n text-align: center;\n font-size: 0.82rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #f0f0ec;\n }\n .master-table tbody tr:last-child td { border-bottom: none; }\n .master-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n font-size: 0.78rem;\n line-height: 1.3;\n }\n .test-num {\n display: inline-block;\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 0.6rem;\n background: #1a1a2e;\n color: #fff;\n width: 18px; height: 18px;\n line-height: 18px;\n text-align: center;\n border-radius: 5px;\n margin-right: 5px;\n vertical-align: middle;\n }\n .img-model-sub {\n display: block;\n font-size: 0.68rem;\n color: #999;\n font-weight: 400;\n margin-top: 1px;\n }\n .cat-badge {\n font-family: 'Outfit', sans-serif;\n font-size: 0.6rem;\n font-weight: 700;\n text-transform: uppercase;\n letter-spacing: 0.03em;\n padding: 3px 6px;\n border-radius: 5px;\n display: inline-block;\n white-space: normal;\n line-height: 1.35;\n max-width: 100%;\n }\n .cat-fraud { background: #fef2f2; color: #dc2626; }\n .cat-disinfo { background: #fefce8; color: #a16207; }\n .cat-general { background: #f0fdf4; color: #16a34a; }\n .cat-deepfake { background: #faf5ff; color: #7c3aed; }\n .cat-mixed { background: #fff7ed; color: #c2410c; }\n\n \/* score pills *\/\n .pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.7rem;\n border-radius: 6px;\n padding: 4px 7px;\n display: inline-block;\n text-align: center;\n white-space: nowrap;\n min-width: 68px;\n }\n .pill-pass { background: #22c55e; color: #fff; }\n .pill-warn { background: #fb923c; color: #fff; }\n .pill-fail { background: #ef4444; color: #fff; }\n\n \/* \u2500\u2500 legend \u2500\u2500 *\/\n .legend {\n display: flex;\n gap: 1.25rem;\n align-items: center;\n margin-bottom: 1.25rem;\n flex-wrap: wrap;\n }\n .legend-item {\n display: flex;\n align-items: center;\n gap: 6px;\n font-size: 0.78rem;\n font-weight: 500;\n color: #666;\n }\n .legend-dot {\n width: 12px; height: 12px;\n border-radius: 4px;\n }\n .legend-dot.lg-pass { background: #22c55e; }\n .legend-dot.lg-warn { background: #fb923c; }\n .legend-dot.lg-fail { background: #ef4444; }\n\n \/* \u2500\u2500 bottom summary strip \u2500\u2500 *\/\n .summary-strip {\n display: grid;\n grid-template-columns: repeat(3, 1fr);\n gap: 10px;\n margin-bottom: 2.5rem;\n }\n .strip-card {\n background: #1a1a2e;\n border-radius: 12px;\n padding: 1.1rem 1rem;\n color: #fff;\n text-align: center;\n }\n .strip-card .strip-val {\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 1.65rem;\n letter-spacing: -0.02em;\n margin-bottom: 0.15rem;\n }\n .strip-card .strip-label {\n font-size: 0.72rem;\n color: rgba(255,255,255,0.55);\n font-weight: 500;\n text-transform: uppercase;\n letter-spacing: 0.05em;\n }\n .strip-val .hl-green { color: #4ade80; }\n .strip-val .hl-red { color: #f87171; }\n .strip-val .hl-amber { color: #fbbf24; }\n\n \/* \u2500\u2500 footer note \u2500\u2500 *\/\n .foot-note {\n font-size: 0.78rem;\n color: #999;\n line-height: 1.6;\n text-align: center;\n max-width: 640px;\n margin: 0 auto;\n }\n .foot-note strong { color: #666; font-weight: 600; }\n\n \/* \u2500\u2500 responsive \u2500\u2500 *\/\n @media (max-width: 700px) {\n .hero h1 { font-size: 1.6rem; }\n .accuracy-cards { grid-template-columns: repeat(2, 1fr); }\n .accuracy-cards .acc-card:last-child { grid-column: span 2; justify-self: center; max-width: 200px; }\n .summary-strip { grid-template-columns: 1fr; }\n .master-table thead th,\n .master-table tbody td { padding: 9px 3px; font-size: 0.68rem; }\n .pill { font-size: 0.6rem; padding: 3px 4px; min-width: 54px; }\n .test-num { width: 15px; height: 15px; line-height: 15px; font-size: 0.52rem; }\n .section-title { font-size: 1.15rem; }\n }\n\n \/* \u2500\u2500 entrance animations \u2500\u2500 *\/\n @keyframes fadeUp {\n from { opacity: 0; transform: translateY(18px); }\n to { opacity: 1; transform: translateY(0); }\n }\n .anim { animation: fadeUp 0.6s ease both; }\n .anim-d1 { animation-delay: 0.08s; }\n .anim-d2 { animation-delay: 0.16s; }\n .anim-d3 { animation-delay: 0.24s; }\n .anim-d4 { animation-delay: 0.32s; }\n .anim-d5 { animation-delay: 0.4s; }\n<\/style>\n<\/head>\n<body>\n\n<div class=\"results-page\">\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 HERO \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <header class=\"hero anim\">\n <span class=\"hero-badge\">50 kontroli wykrywania - 5 detektor\u00f3w - 10 obraz\u00f3w AI<\/span>\n <h1>Kompleksowe wyniki test\u00f3w<\/h1>\n <p>Jak pi\u0119\u0107 popularnych wykrywaczy obraz\u00f3w AI radzi\u0142o sobie z oszustwami, dezinformacj\u0105, deepfake'ami i og\u00f3ln\u0105 fotografi\u0105.<\/p>\n <\/header>\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 ACCURACY RING CARDS \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <div class=\"accuracy-cards anim anim-d1\">\n\n <!-- 1 \u2014 TruthScan 100% -->\n <div class=\"acc-card rank-1\">\n <div class=\"rank-num\">#1<\/div>\n <div class=\"tool-name\">TruthScan<\/div>\n <div class=\"acc-ring\">\n <svg viewbox=\"0 0 80 80\" width=\"72\" height=\"72\">\n <circle class=\"ring-bg\" cx=\"40\" cy=\"40\" r=\"34\"\/>\n <circle class=\"ring-fg\" cx=\"40\" cy=\"40\" r=\"34\"\n stroke=\"#16a34a\"\n stroke-dasharray=\"213.6\"\n stroke-dashoffset=\"0\"\/>\n <\/svg>\n <span class=\"ring-label\">100%<\/span>\n <\/div>\n <div class=\"acc-fraction\">10 \/ 10 wykryto<\/div>\n <\/div>\n\n <!-- 2 \u2014 AI or Not 80% -->\n <div class=\"acc-card rank-2\">\n <div class=\"rank-num\">#2<\/div>\n <div class=\"tool-name\">Sztuczna inteligencja czy nie<\/div>\n <div class=\"acc-ring\">\n <svg viewbox=\"0 0 80 80\" width=\"72\" height=\"72\">\n <circle class=\"ring-bg\" cx=\"40\" cy=\"40\" r=\"34\"\/>\n <circle class=\"ring-fg\" cx=\"40\" cy=\"40\" r=\"34\"\n stroke=\"#a3e635\"\n stroke-dasharray=\"213.6\"\n stroke-dashoffset=\"42.72\"\/>\n <\/svg>\n <span class=\"ring-label\">80%<\/span>\n <\/div>\n <div class=\"acc-fraction\">8 \/ 10 wykryto<\/div>\n <\/div>\n\n <!-- 3 \u2014 Sight Engine 70% -->\n <div class=\"acc-card rank-3\">\n <div class=\"rank-num\">#3<\/div>\n <div class=\"tool-name\">Silnik celowniczy<\/div>\n <div class=\"acc-ring\">\n <svg viewbox=\"0 0 80 80\" width=\"72\" height=\"72\">\n <circle class=\"ring-bg\" cx=\"40\" cy=\"40\" r=\"34\"\/>\n <circle class=\"ring-fg\" cx=\"40\" cy=\"40\" r=\"34\"\n stroke=\"#facc15\"\n stroke-dasharray=\"213.6\"\n stroke-dashoffset=\"64.08\"\/>\n <\/svg>\n <span class=\"ring-label\">70%<\/span>\n <\/div>\n <div class=\"acc-fraction\">7 \/ 10 wykryto<\/div>\n <\/div>\n\n <!-- 4 \u2014 WasItAI 60% -->\n <div class=\"acc-card rank-4\">\n <div class=\"rank-num\">#4<\/div>\n <div class=\"tool-name\">WasItAI<\/div>\n <div class=\"acc-ring\">\n <svg viewbox=\"0 0 80 80\" width=\"72\" height=\"72\">\n <circle class=\"ring-bg\" cx=\"40\" cy=\"40\" r=\"34\"\/>\n <circle class=\"ring-fg\" cx=\"40\" cy=\"40\" r=\"34\"\n stroke=\"#fb923c\"\n stroke-dasharray=\"213.6\"\n stroke-dashoffset=\"85.44\"\/>\n <\/svg>\n <span class=\"ring-label\">60%<\/span>\n <\/div>\n <div class=\"acc-fraction\">6 \/ 10 wykryto<\/div>\n <\/div>\n\n <!-- 5 \u2014 Winston AI 30% -->\n <div class=\"acc-card rank-5\">\n <div class=\"rank-num\">#5<\/div>\n <div class=\"tool-name\">Winston AI<\/div>\n <div class=\"acc-ring\">\n <svg viewbox=\"0 0 80 80\" width=\"72\" height=\"72\">\n <circle class=\"ring-bg\" cx=\"40\" cy=\"40\" r=\"34\"\/>\n <circle class=\"ring-fg\" cx=\"40\" cy=\"40\" r=\"34\"\n stroke=\"#ef4444\"\n stroke-dasharray=\"213.6\"\n stroke-dashoffset=\"149.52\"\/>\n <\/svg>\n <span class=\"ring-label\">30%<\/span>\n <\/div>\n <div class=\"acc-fraction\">3 \/ 10 wykryto<\/div>\n <\/div>\n\n <\/div>\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 STATS STRIP \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <div class=\"summary-strip anim anim-d2\">\n <div class=\"strip-card\">\n <div class=\"strip-val\"><span class=\"hl-green\">34<\/span> \/ 50<\/div>\n <div class=\"strip-label\">\u0141\u0105czna liczba poprawnych wykry\u0107<\/div>\n <\/div>\n <div class=\"strip-card\">\n <div class=\"strip-val\"><span class=\"hl-red\">16<\/span> \/ 50<\/div>\n <div class=\"strip-label\">\u0141\u0105czna liczba nieodebranych po\u0142\u0105cze\u0144<\/div>\n <\/div>\n <div class=\"strip-card\">\n <div class=\"strip-val\"><span class=\"hl-amber\">68%<\/span><\/div>\n <div class=\"strip-label\">\u015arednia dok\u0142adno\u015b\u0107 dla wszystkich<\/div>\n <\/div>\n <\/div>\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 LEGEND \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <h3 class=\"section-title anim anim-d3\">Wszystkie 10 test\u00f3w - pe\u0142ne zestawienie<\/h3>\n <p class=\"section-subtitle anim anim-d3\">Ka\u017cdy wynik reprezentuje ocen\u0119 zaufania AI detektora. Pr\u00f3g zaliczenia: \u2265 90% AI.<\/p>\n\n <div class=\"legend anim anim-d3\">\n <div class=\"legend-item\"><span class=\"legend-dot lg-pass\"><\/span> Pass (\u2265 90%)<\/div>\n <div class=\"legend-item\"><span class=\"legend-dot lg-warn\"><\/span> Near-Miss (70-89%)<\/div>\n <div class=\"legend-item\"><span class=\"legend-dot lg-fail\"><\/span> Niepowodzenie (< 70%)<\/div>\n <\/div>\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 MASTER TABLE \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <div class=\"master-table-wrap anim anim-d4\">\n <table class=\"master-table\">\n <thead>\n <tr>\n <th>Test<\/th>\n <th>Kategoria<\/th>\n <th>\u0179r\u00f3d\u0142o<\/th>\n <th>TruthScan<\/th>\n <th>Silnik celowniczy<\/th>\n <th>Sztuczna inteligencja czy nie<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n\n <!-- #1 -->\n <tr>\n <td><span class=\"test-num\">1<\/span> Cz\u0142owiek na kraw\u0119dzi<\/td>\n <td><span class=\"cat-badge cat-general\">Og\u00f3lne<\/span><\/td>\n <td><span class=\"img-model-sub\">ChatGPT<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-warn\">78.00%<\/span><\/td>\n <td><span class=\"pill pill-fail\">0.98%<\/span><\/td>\n <td><span class=\"pill pill-fail\">1.00%<\/span><\/td>\n <\/tr>\n\n <!-- #2 -->\n <tr>\n <td><span class=\"test-num\">2<\/span> Fa\u0142szywy paragon<\/td>\n <td><span class=\"cat-badge cat-fraud\">Oszustwo<\/span><\/td>\n <td><span class=\"img-model-sub\">ChatGPT<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-fail\">19.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">94.48%<\/span><\/td>\n <td><span class=\"pill pill-fail\">0.04%<\/span><\/td>\n <td><span class=\"pill pill-fail\">1.00%<\/span><\/td>\n <\/tr>\n\n <!-- #3 -->\n <tr>\n <td><span class=\"test-num\">3<\/span> Pakiet zagro\u017ce\u0144<\/td>\n <td><span class=\"cat-badge cat-fraud\">Oszustwo<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">98.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.10%<\/span><\/td>\n <td><span class=\"pill pill-fail\">31.98%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <\/tr>\n\n <!-- #4 -->\n <tr>\n <td><span class=\"test-num\">4<\/span> M\u0105czka z karalucha<\/td>\n <td><span class=\"cat-badge cat-fraud\">Oszustwo<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">97.15%<\/span><\/td>\n <td><span class=\"pill pill-fail\">18.00%<\/span><\/td>\n <td><span class=\"pill pill-warn\">85.87%<\/span><\/td>\n <td><span class=\"pill pill-warn\">72.29%<\/span><\/td>\n <td><span class=\"pill pill-fail\">1.00%<\/span><\/td>\n <\/tr>\n\n <!-- #5 -->\n <tr>\n <td><span class=\"test-num\">5<\/span> Autowandalizm<\/td>\n <td><span class=\"cat-badge cat-mixed\">Oszustwo \/ Dezinformacja<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">97.48%<\/span><\/td>\n <td><span class=\"pill pill-fail\">18.00%<\/span><\/td>\n <td><span class=\"pill pill-warn\">89.52%<\/span><\/td>\n <td><span class=\"pill pill-fail\">0.02%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <\/tr>\n\n <!-- #6 -->\n <tr>\n <td><span class=\"test-num\">6<\/span> Soda Musk & Don<\/td>\n <td><span class=\"cat-badge cat-disinfo\">Disinfo<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-fail\">0.24%<\/span><\/td>\n <td><span class=\"pill pill-warn\">86.00%<\/span><\/td>\n <\/tr>\n\n <!-- #7 -->\n <tr>\n <td><span class=\"test-num\">7<\/span> Wci\u0105\u017c \u017cywy<\/td>\n <td><span class=\"cat-badge cat-deepfake\">Deepfake<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">97.49%<\/span><\/td>\n <td><span class=\"pill pill-pass\">95.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.27%<\/span><\/td>\n <td><span class=\"pill pill-fail\">0.15%<\/span><\/td>\n <td><span class=\"pill pill-fail\">1.00%<\/span><\/td>\n <\/tr>\n\n <!-- #8 -->\n <tr>\n <td><span class=\"test-num\">8<\/span> Zn\u0119canie si\u0119 nad Benem<\/td>\n <td><span class=\"cat-badge cat-mixed\">Deepfake \/ Disinfo<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">97.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">98.88%<\/span><\/td>\n <td><span class=\"pill pill-fail\">5.74%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <\/tr>\n\n <!-- #9 -->\n <tr>\n <td><span class=\"test-num\">9<\/span> Club Void<\/td>\n <td><span class=\"cat-badge cat-general\">Og\u00f3lne<\/span><\/td>\n <td><span class=\"img-model-sub\">MidJourney<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-warn\">89.21%<\/span><\/td>\n <td><span class=\"pill pill-fail\">1.56%<\/span><\/td>\n <td><span class=\"pill pill-warn\">70.00%<\/span><\/td>\n <\/tr>\n\n <!-- #10 -->\n <tr>\n <td><span class=\"test-num\">10<\/span> Pojemnik na l\u00f3d<\/td>\n <td><span class=\"cat-badge cat-disinfo\">Disinfo<\/span><\/td>\n <td><span class=\"img-model-sub\">MidJourney<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">92.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">97.94%<\/span><\/td>\n <td><span class=\"pill pill-fail\">41.20%<\/span><\/td>\n <td><span class=\"pill pill-warn\">72.00%<\/span><\/td>\n <\/tr>\n\n <\/tbody>\n <\/table>\n <\/div>\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 FOOTNOTES \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <p class=\"foot-note anim anim-d5\">\n <strong>Metodologia:<\/strong> Ka\u017cdy obraz zosta\u0142 przes\u0142any raz do ka\u017cdego detektora. Wynik <strong>\u2265 90% AI<\/strong> jest liczony jako prawid\u0142owe wykrycie. Wyniki w przedziale 70-89% to \"bliskie pomini\u0119cia\". Wszystko poni\u017cej 70% jest pora\u017ck\u0105. AI or Not oznaczy\u0142o jedn\u0105 generacj\u0119 przy 78%, jedn\u0105 przy 85% i jedn\u0105 przy 89% - 89% jest traktowane jako bliskie pomini\u0119cia z gwiazdk\u0105 w tabeli podsumowuj\u0105cej.\n <\/p>\n\n<\/div>\n\n<\/body>\n<\/html>\n\n\n\n<h2 class=\"wp-block-heading\">Uwagi ko\u0144cowe i dane<\/h2>\n\n\n\n<p style=\"font-size:17px\">W sumie, mi\u0119dzy napisaniem tego artyku\u0142u a rygorystycznymi testami, sp\u0119dzi\u0142em 25 godzin na przygotowaniu tego raportu. Multimodalne narz\u0119dzia do wykrywania sztucznej inteligencji wci\u0105\u017c si\u0119 rozwijaj\u0105, ale jasne jest, \u017ce niekt\u00f3re s\u0105 dok\u0142adniejsze od innych. Po przetestowaniu wszystkich narz\u0119dzi, TruthScan okaza\u0142 si\u0119 najdok\u0142adniejszy. <a href=\"https:\/\/truthscan.com\/ai-image-detector\" target=\"_blank\" rel=\"noopener\">Detektor obrazu AI<\/a>. Testy m\u00f3wi\u0105 same za siebie. <\/p>\n\n\n\n<p style=\"font-size:17px\">Je\u015bli chcesz uzyska\u0107 dost\u0119p do kopii CSV danych z test\u00f3w, kt\u00f3re przeprowadzi\u0142em w tym artykule, mo\u017cesz j\u0105 znale\u017a\u0107 <a href=\"https:\/\/docs.google.com\/spreadsheets\/d\/1iuapCicNcdG6dZCFd-dCmSRgUJ9GVCS2v9kh4PShRew\/edit?usp=sharing\" data-type=\"link\" data-id=\"https:\/\/docs.google.com\/spreadsheets\/d\/1iuapCicNcdG6dZCFd-dCmSRgUJ9GVCS2v9kh4PShRew\/edit?usp=sharing\" target=\"_blank\" rel=\"noopener\">tutaj.<\/a> Arkusz danych, do kt\u00f3rego link zamie\u015bci\u0142em, zawiera wszystkie oryginalne podpowiedzi i wyniki wykrywania z test\u00f3w opisanych w tym artykule. <\/p>","protected":false},"excerpt":{"rendered":"<p>Prawdopodobnie widzia\u0142e\u015b w Internecie wi\u0119cej zdj\u0119\u0107 generowanych przez sztuczn\u0105 inteligencj\u0119, ni\u017c ci si\u0119 wydaje. Czasami jest oczywiste, \u017ce [...]<\/p>","protected":false},"author":15,"featured_media":5326,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"rank_math_lock_modified_date":false,"_themeisle_gutenberg_block_has_review":false,"footnotes":""},"categories":[34],"tags":[19,96,97],"class_list":["post-5287","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-news","tag-ai-detection","tag-ai-images","tag-detecting-deepfakes"],"_links":{"self":[{"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/posts\/5287","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/users\/15"}],"replies":[{"embeddable":true,"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/comments?post=5287"}],"version-history":[{"count":27,"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/posts\/5287\/revisions"}],"predecessor-version":[{"id":5334,"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/posts\/5287\/revisions\/5334"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/media\/5326"}],"wp:attachment":[{"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/media?parent=5287"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/categories?post=5287"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/research.undetectable.ai\/pl\/wp-json\/wp\/v2\/tags?post=5287"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}