← Back to live feed1 story
OpenAI is restricting access to its EU watermark detection tool, limiting it to vetted researchers and expert organizations due to vulnerabilities in its invisible markers. The textGrain system will be deployed for ChatGPT and Codex users in the European Union over the coming weeks to comply with Article 50 of the EU AI Act, with a global opt-in available for API customers today.
Sign in to suggest edits
Key sources
- SOURCE@openai“In the EU, we’ll start watermarking eligible text from ChatGPT and Codex over the coming weeks to comply with the EU AI Act.”x.com
- SOURCE@openai“Rewriting or translating text can completely remove the watermark.”x.com
- SUPPORT@hesamation“we’re getting ChatGPT watermarks in EU only.”x.com
- SUPPORT@andrewcurran_“OpenAI (for now) is launching their watermarking - textGrain - 𝘰𝘯𝘭𝘺 in the EU, not globally.”x.com
- SUPPORT@kimmonismus“replacing 25% of words with synonyms cut watermark detection from about 92% to 17%.”x.com
- SUPPORT@rohanpaul_ai“swapping 10% of words for synonyms in 400-token passages cut detection from about 92% to 66%, and swapping 25% cut it to 17%.”x.com
- SUPPORT@kunchenguid“a human producing a piece of complete slop using LLM alone, vs a human spending hours and hours crafting their idea but let LLM handle some of the final touch, will both be labeled as "AI-generated"”x.com
- SUPPORT@max_spero_“they evaluate it at a target false positive rate of ~1%!”x.com