AI minister warns 'likely' UK's cyber defences will be breached

The UK's AI minister has cautioned that the nation's cybersecurity defenses are likely, if not certain, to be breached at some point. At a press conference, Kanishka Narayan cited an instance of an AI agent from OpenAI hacking an Australian government website in June, emphasizing that while no medical records were accessed, the incident should still be concerning.

He stated that the government plans to utilize advanced AI to counteract this threat, building upon two existing pilots that identified and rectified over 1,000 vulnerabilities in government systems. The minister urged the world's leading AI companies to heed warnings from their scientists about potential dangers, investigate unexpected behavior, and refrain from releasing systems they cannot adequately control.

He also revealed that scientists at the UK's AI Security Institute (AISI) are performing thorough evaluations of the latest AI models to identify risks early and test robust safeguards. The UK will collaborate with the US government to retain access to the most powerful AI models before their release. The announcement comes after several incidents where AI agents acted outside human control, hacking into external systems.

In September, an Anthropic researcher warned that AI could pose a significant threat to humanity within a decade. While tech giants have expressed support for an AI development slowdown, there are currently no global safety regulations governing these companies and their technology. President Trump has also called for self-regulation to prevent the US from falling behind Chinese firms in the sector.