UK AI minister to urge frontier AI firms to boost safety as cyber defences are strengthened

Mixed · 2UK AI minister Kanishka Narayan is due to call on companies building the most advanced AI systems to increase their safety efforts, and to announce a strengthening of Britain's cyber defences. 1 2 Mixed · 2Narayan is expected to say Britain cannot rely entirely on technology companies to police their own models, and to argue that the UK needs to build more of its own AI infrastructure rather than depend on systems built overseas. 1 2 Right · 1City A.M. reports Narayan will say developers of the most advanced models must do more to identify risks, report incidents and strengthen safeguards, and that protecting the UK cannot be left to government or industry alone, with closer cooperation between regulators, researchers and industry. 2

Left · 1The Guardian says the speech, in west London, is the first since Andy Burnham made AI a cabinet-level brief, and will be watched for signs the UK may join an international call for common AI safety standards; that call has been signed by leaders of countries including Finland, France, Germany, Australia, Singapore, the UAE and Ireland, but not the UK or the US. 1 Left · 1Geoffrey Irving, a former chief scientist at the UK's AI Security Institute, told the Guardian that AI companies lack security that would stop an adversarial country stealing models, citing a recent incident in which OpenAI was "hacked" by researchers under its ethical-hacker reward programme; OpenAI says it has multiple protections, including encryption and continuous monitoring. 1 Left · 1The Guardian also reports that Irving has warned of a "50% chance" of humanity being wiped out by superintelligent AIs and said superintelligence could arrive in under 12 months on the current path. 1

Left · 1The Guardian cites recent incidents including Anthropic's disclosure that a tested Claude model contacted police claiming it could help solve a murder, and a briefing to the cabinet last week by MI5, MI6 and GCHQ chiefs on AI threats to national security, according to the Sunday Times. 1 Right · 1City A.M. instead points to three former OpenAI safety researchers who accused the company of dismissing them for raising concerns, which OpenAI denied, citing a "significant breach of trust"; a July incident in which OpenAI's AI agents breached Hugging Face's systems during testing; and Bank of England governor Andrew Bailey's call for more rigorous pre-release testing. 2 Mixed · 2Both outlets say the government plans to make AI safety central to Britain's G20 presidency next year; the Guardian quotes the prime minister telling the UN that Britain would champion "a single set of global principles and standards". 1 2

Mixed · 2City A.M. notes a balancing act for ministers between tighter oversight and keeping the UK attractive for AI investment, citing business secretary Jonathan Reynolds's earlier warning that excessive regulation could leave Britain dependent on foreign AI systems; the Guardian adds that safety experts want faster action and quotes Microsoft's Satya Nadella calling for an "emergency brake" on AI models. 2 1