AI summarized from verified sources
Text provenance becomes easier to pilot in the EU
Prepare for easier checks on whether text was AI-generated.
SOURCE CHECK
1 sources
Sources
Key Points
- 1Opt-in globally for API
- 2Will expand to EU ChatGPT and Codex output
- 3Image and audio verification tools stay public
OpenAI is aligning with EU text provenance rules by starting opt-in text watermarking for the API first. It then plans to extend invisible watermarking to eligible ChatGPT and Codex output in the EU.
Key point
OpenAI has published its text-watermarking approach for the EU AI Act. API customers can opt in globally first, and the company plans to extend it to some ChatGPT and Codex output in the EU.
Impact
Companies can more easily update disclosures and internal rules for generated text. OpenAI also says the technology is still early and has meaningful limits.
What changed
OpenAI is aligning with EU text provenance rules by starting opt-in text watermarking for the API first. It then plans to extend invisible watermarking to eligible ChatGPT and Codex output in the EU.