OpenAI admits text watermarks are fragile — detector restricted to approved researchers
OpenAI · x · 2026-10-06
OpenAI acknowledges the limits of its text watermarking:
- Watermarks are often undetectable, especially in short passages; rewriting or translating text can remove them entirely
- For now, only approved researchers get access to the detector to help evaluate and improve the technology
- Text watermarking is treated as an ongoing research area, with continued testing based on feedback from users, developers, policymakers, and researchers
Related event: OpenAI launches textGrain text watermarking, starting in the EU(25 posts)→
More from Safety
- Google pushing OEMs against GrapheneOS and bootloader unlocking, project says — mraniki · 2026-11-20
- OpenAI Foundation board member Paul Christiano warns AI takeover could unfold in months — EvanHub · 2026-11-20
- GitHub adds reachability analysis to Dependabot alerts to prioritize real vulnerability risk — 0xkarasy · 2026-11-20
- India launches PrAImaan, a national platform for evaluating healthcare AI, targeting 1,000+ clinicians — DrDatta_AIIMS · 2026-11-20
- US rules poised to block Chinese AI chips made by TSMC and Samsung; Alibaba already moved to SMIC — terryyuezhuo · 2026-11-20
- Bridgewater chart: AI-enabled cyber incidents surge exponentially in recent months — GaryMarcus · 2026-11-20