The UK government will expand a programme that uses frontier AI models to find vulnerabilities in its own systems. A pilot found 407 weaknesses across nine government organisations in a month for £13,000 in tokens. The government is also considering legislation on AI incident reporting and loss-of-control risk.
