UK to tighten cyber defences and press AI firms on safety

The United Kingdom is preparing to reinforce its cyber defences and push companies building the most powerful AI systems to do more on model safety. AI Minister Kanishka Narayan is expected to make the case in a speech in west London, arguing that Britain cannot depend entirely on technology companies to police their own models.

Narayan is also expected to support development of domestic British AI infrastructure. The speech comes as international AI safety standards are being discussed with countries including Finland, France, Germany, Australia, Singapore, the United Arab Emirates and Ireland, and after AI became a cabinet-level area of responsibility.

Security concerns have intensified around the risk that hostile states could steal undisclosed advanced models. Geoffrey Irving, former chief scientist at the UK’s AI Safety Institute, said company protections may not be enough, citing an OpenAI test in which cybersecurity researchers reported broad potential access under an ethical hackers programme. OpenAI said it uses several layers of model protection, including encryption and continuous monitoring.

UK intelligence chiefs recently briefed the cabinet on AI-related national security threats. The prime minister also told the UN that Britain would use its G20 presidency next year to promote common global principles and AI safety standards.