Anthropic said on Thursday it has ​granted a team from nonprofit ‌AI evaluator METR employee-level access to its systems, allowing the group to independently ​assess its advanced models, safety ​practices and risk management.