UK AI Security Institute
Follow how frontier capabilities and risks are tested by a public institution.
Context and limitations
Disclosed evaluations cover selected systems and threat models; missing public results are not a safety finding.
Public research and evaluation explanations with linked artifacts