Anthropic gives outside AI watchdog “employee-like” access (arstechnica.com)

The outside watchdog can publish its findings without Anthropic's approval.