Artificial Intelligence: Safety
The question
To ask the Minister for the Cabinet Office, with reference to the Answer of 28 April 2026 to Question 128813, whether the AI Security Institute has identified vulnerabilities in models deployed before the Institute confirmed those vulnerabilities had been resolved.
Answered by Department for Business, Innovation, Science and Trade
The AI Security Institute is a research institute, not a regulator. Its scientific testing and research help government, developers and existing regulators understand the risks posed by advanced AI and inform any action needed.
AISI has identified dozens of AI model vulnerabilities that leading frontier AI developers have addressed before releasing them to the public, protecting British citizens from potential harm.
The Institute tests models both before and after deployment. Where testing identifies a vulnerability, findings are shared with the developer at pace so that mitigations can be put in place, regardless of deployment status.
Verbatim has judged this answer against the question that was actually asked — answered, partly answered, or evaded. Sign in to see the verdict →