METR releases frontier risk report assessing agent loss-of-control risks at Anthropic, Google, Meta, and OpenAI
Original title:Frontier Risk Report (February to March 2026)
AI summary
METR conducted a pilot assessment from February 16 to March 16, 2026, with Anthropic, Google, Meta, and OpenAI participating and sharing their strongest internal models and non-public information.
Key Takeaways
- METR conducted a pilot assessment from February 16 to March 16
- with Anthropic
- Google,and OpenAI participating and sharing their strongest internal models and non-public…
Quick Facts
- Published
- Updated
- Source
- METR:Research(网页)
- Category
- Papers & Research
- AI score
- 76
METR 联合四家前沿 AI 公司评估内部 Agent 的失控风险,给出手段、动机、机会三分框架和量化基准结果,可帮助读者理解内部部署风险的现状边界。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: METR releases frontier risk report assessing agent loss-of-control risks at Anthropic, Google, Meta, and OpenAI. May 19, 2026. https://new.nat6.net/en/news/cmu1pnjrn05pzrodupi9h6ta3/