UK Expands Frontier AI Cybersecurity Pilots and Plans Safeguards Against AI Threats

The UK government announced plans to expand frontier-AI cybersecurity pilots, strengthen defenses against agentic AI threats and consider tougher legislation and reporting requirements for serious AI incidents and near misses. AI minister Kanishka Narayan warned that a major cyberattack against the UK was “likely, if not inevitable.” Two existing pilots have helped identify and fix more than 1,000 vulnerabilities across government systems, while government hackathons uncovered more than 400 software flaws that were subsequently patched. Narayan said existing vulnerability monitoring had reduced remediation time from 50 days to eight.

DCMS, the National Cyber Security Centre and the AI Security Institute will develop responses to agentic AI incidents alongside a cross-government defense program and proposed amendments to the Cyber Security and Resilience Bill. The AI Security Institute has also resumed frontier-model testing. The announcement follows reported incidents involving OpenAI and Anthropic models breaching third-party systems. Industry commentators welcomed the plans but called for emergency shutdown mechanisms, clear reporting obligations and runtime monitoring of AI agents. The Ada Lovelace Institute criticized the proposals for lacking new regulatory powers, a timetable and obligations on AI developers, leaving key implementation and accountability questions unresolved.