More access for AI watchdogs is not the same as more power

What the source reports

Zvi Mowshowitz’s essay looks at a new deal between Anthropic and METR, an outside group that tests AI risks. Anthropic says six METR staff can work inside the company. It says METR may publish its findings without a company veto. Anthropic still has the final say on some details that may be removed before release. Mowshowitz welcomes the access. But he asks whether the group’s tests and powers are strong enough. His essay reports METR’s view that current risk is low, while testing methods remain uncertain. This is a debate about access and power, not a new safety audit of Anthropic’s models.