Swap 1 word in 10 and the AI watermark's gone.

OpenAI is adding an invisible watermark called textGrain to ChatGPT and Codex text in the EU to comply with the EU AI Act. It survives copy-paste because it lives in the model's word choices, not hid…

Source: OpenAI, 'Our approach to EU text provenance rules' (Oct 5 2026) + technical report 'textGrain: Entropy-Calibrated Watermarking for Language Model Text' (Oct 5 2026; authors UPenn + Yale + OpenAI). Verified mirror: TechCrunch, 'OpenAI will start watermarking ChatGPT's text in the EU', Aditya Mehta, Oct 5 2026 (https://techcrunch.com/2026/10/05/openai-will-start-watermarking-chatgpts-text-in-the-eu/); BeInCrypto, Oct 6 2026. Facts: EU AI Act Article 50 text-transparency obligations apply from Aug 2 2026; OpenAI adds an invisible textGrain watermark to eligible ChatGPT and Codex text in the EU over coming weeks (all plans, EU only, not a global default); API opt-in worldwide, off by default; detector access limited to approved researchers/expert orgs; open source planned. Detection: ~66% of unedited ~150-word answers, ~92% at ~300 words; swap 1 word in 10 -> 66%; swap 1 in 4 -> 17%. Watermark adds no hidden characters. Missing watermark does not prove human authorship.

More memes on lilmeme.fun

Open on lilmeme.fun