Third-party AI safety assessments
AssessTechniques
Independent external evaluations of AI systems for safety and security risks.
Why it's here
Placed in Assess: 3 article(s) of evidence from 2 source(s), led by framework updates, with 1 in the last 30 days. Confidence 49%.
Evidence (3)
- 7OpenAI Blog·9/22/2026framework_updateOpenAI sets principles for third-party AI safety assessments
OpenAI outlines priorities and principles for rigorous, secure, and independent third-party assessments of frontier AI models and their safeguards. The guidance emphasizes evaluation quality, independence, and security as part of responsible model oversight.
- 7The New Stack·7/2/2026framework_updateFedCM aims to replace third-party-cookie social login
The article explains that legacy social login flows such as "Sign in with Google" depend on third-party cookies and are increasingly fragile as browsers block or phase them out. It presents FedCM (Federated Credential Management) as a browser-mediated standard that preserves federated sign-in without cross-site tracking, using a native consent prompt and explicit user approval.
- 5OpenAI Blog·5/29/2026framework_updateOpenAI publishes a playbook for third-party AI evaluations
OpenAI shared guidance for conducting third-party evaluations of frontier AI systems. The playbook covers how to assess model capabilities, safety safeguards, and the validity of evaluation methods.