Frontier Risk Report (February to March 2026) — METR

What it is

METR's pilot assessment of rogue deployment risk at frontier AI companies: an exercise assessing misalignment risks from AI agents used inside frontier AI developers, with participation from Anthropic, Google, Meta and OpenAI.

Topics