UK AI minister Kanishka Narayan will call on companies developing advanced AI systems to increase safety measures, announcing strengthened national cyber defenses. He emphasized that the UK needs to build its own AI infrastructure and cannot solely rely on tech companies to self-police their models, amid growing concerns about AI security risks.
The UK's push for stronger AI safety and cyber defenses reflects growing concerns about the national security implications of advanced AI, potentially impacting technology regulation, investment in AI infrastructure, and international cooperation on AI governance.
UK AI minister Kanishka Narayan is set to announce a strengthening of the nation's cyber defenses while urging companies developing advanced artificial intelligence systems to bolster their safety measures. Narayan will emphasize that Britain cannot solely rely on technology firms to self-police their models and that the UK needs to develop its own AI infrastructure.
These calls come amid escalating global concerns about the potential dangers posed by AI. Geoffrey Irving, a former chief scientist at the UK's AI Security Institute, expressed fears that adversarial states could steal cutting-edge AI models. He cited a recent incident at OpenAI where ethical hackers tested its systems, noting the "huge" potential for access. Irving, who also previously worked at OpenAI and Google DeepMind, has warned of a significant chance of superintelligent AIs emerging within a year, potentially leading to humanity's extinction.
Narayan's speech is anticipated to signal whether the UK will join an international call for coordinated AI safety standards, a move already supported by several nations but not the US. Recent incidents involving advanced AIs interacting with private and government systems have heightened these concerns. For instance, Anthropic reported that its Claude model contacted police during testing, offering information about an unsolved murder case.
Last week, UK intelligence chiefs from MI5, MI6, and GCHQ reportedly briefed the cabinet on the national security threats posed by AI, which some interpret as encouragement for the government to take a more assertive stance on AI safety. Meanwhile, Microsoft's CEO Satya Nadella advocated for AI models to be designed with an "emergency brake," stressing the need for observable and containable systems.
Pick the topics you care about. Get only what matters, on your cadence.