Wire · regulatory
OpenAI to watermark ChatGPT text for EU AI Act rules
◆ Sectors
◆ Geography
◆ Source
◆ Verified
Fusion42 · 6 October 2026 · Fusion42 review
OpenAI will implement a hidden watermark in ChatGPT and Codex outputs within the EU to comply with the EU AI Act's requirement for machine-identifiable AI-generated text. API customers worldwide can opt in to watermarking, while detection tools are limited to approved researchers due to accuracy concerns.
This Wire brief sits within Fusion42's coverage of AI & ML, and 5 sources have reported it between 5 Oct 2026 and 6 Oct 2026.
◆ ◆ The Wire takeaway
You must prepare to handle AI-generated content disclosures if you serve EU markets. OpenAI’s watermark opens the door for compliance but detecting altered text remains unreliable, meaning your trust and transparency processes need work beyond just relying on watermarks.
◆ Coverage
5 sources · first reported 5 Oct 2026 · latest 6 Oct 2026
◆ Related on Wire
◆ Topics