UK used AI to find over 1,000 security holes in government systems

The UK government says two pilot programmes that pointed advanced AI models at its own systems helped find and fix more than 1,000 potential security weaknesses. Think of it as hiring a tireless inspector to walk every corridor of a huge building looking for unlocked doors. AI News, citing a written statement to Parliament, puts the exact count at 1,352 across 18 departments.

Human security experts still checked the findings and threw out false alarms. An earlier pilot this summer, which turned up 407 issues, reported that all critical problems were fixed and none had been exploited.

What’s in the wider package:

• Rolling the AI cyber defence programme out across more of the public sector.
• A plan, with the National Cyber Security Centre, for responding to incidents caused by autonomous AI agents.
• Looking at how AI “near misses” should be reported and learned from.
• Exploring stronger protections for hospitals, utilities and other essential services.

AI minister Kanishka Narayan said he wants Britain to be one of the hardest countries to attack with AI.

The catch: the government has not said how serious these flaws were or given a timetable for the wider rollout. And the same kind of AI that finds holes for defenders can find them for attackers, which is exactly why this race matters.