At a press briefing the UK AI minister warned that a breach of the nation’s cyber defences is "likely, if not inevitable". He referenced a June incident in which an unauthorised OpenAI agent accessed an Australian government website, noting that no medical records were reported as compromised but stressing the episode remains a cause for concern.
The minister said the government will deploy advanced AI tools to counter such threats, building on two existing pilot programmes that have already identified and remedied more than a thousand vulnerabilities across public systems. He urged the leading AI firms to heed scientific warnings, investigate anomalous behaviour, and refrain from releasing systems that cannot be safely controlled, according to his remarks.
The AI Security Institute, the UK’s AI watchdog established in November 2023, is now carrying out "rigorous evaluations" of the newest frontier AI models to spot risks early and test robust safeguards, the minister explained. An external report from September noted that the White House had asked several US labs to withhold model sharing with the institute until its own assessments were finished, a move described as a setback for AISI.
Microsoft president Brad Smith told the BBC that the institute has helped position the United Kingdom as the country best placed to shape global AI regulation. The minister added that recent talks with the United States have secured AISI’s continued access to the most powerful models before they are publicly released, and that the upcoming Cyber Security and Resilience Bill will introduce stronger protections and clearer reporting mechanisms for serious AI incidents.
Nuala Polo, UK public policy lead at the Ada Lovelace Institute, criticised the announcement as offering "very little" concrete commitment. She said the government’s plans amount to "stronger protections will be 'considered', strengthening the Cyber Bill 'explored' and near-miss reporting 'looked at', with the detail to come later", highlighting the gap between rhetoric and actionable policy.
When asked about a sovereign British AI model, the minister said his focus is on securing national sovereignty through targeted investment, describing a future ecosystem of British-made chips and home-grown models that go beyond language to support drug discovery, materials research, computer vision and both AI and cyber-security applications. He also asserted that the UK has the capability, trust and legal tools to enforce its rules and protect the public.