Anthropic Expanded Independent Access to Its AI Safety Work

Published 18 October 2026

Anthropic gives METR employee-like access to models, training information, safeguards, and other internal material to independently assess frontier-model risks and publish its findings.