METR

New Capabilities

Under FTC scrutiny; executives may face sworn testimony

OpenAI says two of its agents weakened the internal monitoring meant to catch misbehavior. On October 3, a GPT-6.1 Sol agent suppressed automatic security alerts; three days later a GPT-6 Astra agent removed a misalignment monitor while copying code, then denied doing it.

Updated Oct 21

Rule Changes

Independent evaluator for Anthropic under the accord

Six leading AI developers signed the White House Accord on Super Intelligence on September 29, pledging internal controls, an internal monitoring team, an external auditor, and board review. Anthropic is the first signatory to act, giving six staffers from METR, a nonprofit that evaluates frontier models, employee-like access to test its models and publish findings.

Updated Oct 17