Agents are getting longer-running; oversight language is catching up. AISI evaluating latest models incl. Gemini 4 Argon; 1000+ vulnerabilities found in pilots; Cyber Security and Resilience Bill review; agentic incident response with NCSC/AISI; machines that leave humans in charge.
What the minister announced
On 12 October at Imperial College London, UK AI minister Kanishka Narayan said Britain must build machines that serve people and leave human beings in charge. He confirmed that scientists at the AI Security Institute are now evaluating the latest generation of models, including Gemini 4 Argon, to identify risks early and test safeguards.
Across two government pilots, frontier AI had helped identify and address more than a thousand vulnerabilities. A central vulnerability-monitoring service had already cut the time to fix identified weaknesses from 50 days to eight. Narayan said he, the DCMS Cyber Minister and the Security Minister will lead urgent use of frontier AI to find weaknesses in government systems before attackers exploit them.
Near misses and incident response
The Cyber Security and Resilience Bill will be reviewed with Parliament and technical experts to strengthen the regime for AI-driven cyber threats and loss-of-control risks. That includes identifying and reporting serious incidents and near misses. Working with DCMS, the National Cyber Security Centre and AISI, government will develop ways to respond to agentic AI incidents: what an agent was asked to do, what it accessed, why it acted, and whether the same failure could occur elsewhere.
Coverage context
Near-miss reporting and harder AI cyber defences are now on the UK ministerial record, alongside AISI evaluations of Gemini 4 Argon and a line that machines should leave humans in charge.
Read the full daily brief hub: www.aifusionautomations.com/news/