Third-Party AI Assessment

The practice of having parties outside an AI developer test its models, safeguards and safety arguments, together with the access levels, scope agreements and publication terms that such testing depends on.