UK AI minister Kanishka Narayan calls for countries to harden defenses against AI risks

UK AI Minister Kanishka Narayan has called on nations to harden their defenses against artificial intelligence, warning that model testing is no longer sufficient to mitigate growing risks. Speaking at the Labour Party conference, Narayan highlighted immediate threats like cybersecurity and biochemical risks, while also prioritizing contingency plans for potential, unprecedented job losses. Think tank IPPR warned that a worst-case scenario could negatively affect 8 million UK jobs, with women and service workers facing the highest exposure. The UK plans to coordinate global safety principles during its G20 presidency next year, despite US President Donald Trump warning that slowing research would benefit China.

Narayan on hardening AI defenses

• ▪UK AI Minister Kanishka Narayan stated on September 28, 2026, that countries must harden and build their defenses against AI risks, calling existing model testing insufficient.
• ▪UK AI Minister Kanishka Narayan stated on September 28, 2026, that his primary focus is on current risks, including cybersecurity, biochemical risks, and models exceeding human instructions.

Using G20 presidency for AI governance

• ▪UK Foreign Secretary Ed Miliband stated on September 28, 2026, that the UK will use its G20 presidency next year to lead efforts to manage AI safety risks.
• ▪UK AI Minister Kanishka Narayan stated on September 28, 2026, that the UK intends to use its G20 presidency next year to convene countries and shape global AI principles.

Narayan's contingency plans for AI job losses

• ▪UK AI Minister Kanishka Narayan stated on September 28, 2026, that contingency measures for severe AI job losses could involve changes to labour market policy and regulation.
• ▪UK AI Minister Kanishka Narayan stated on September 28, 2026, that one of his top priorities is establishing a contingency plan for worst-case AI job loss scenarios.

AI and the workforce

• ▪UK AI Minister Kanishka Narayan promised on September 28, 2026, that workers will have a voice and there will be better transparency when artificial intelligence is adopted.
• ▪UK AI Minister Kanishka Narayan stated on September 28, 2026, that governments, companies, and individuals have agency in determining whether AI augments or automates human labor.

IPPR predictions on AI labor impact

• ▪Carsten Jung of the IPPR think tank stated on September 28, 2026, that women are 40% more exposed than men to the threat of AI due to a higher proportion working in service jobs.
• ▪Carsten Jung of the IPPR think tank stated on September 28, 2026, that a rise of sophisticated agentic AI could increase the proportion of exposed tasks to about 60% across the UK economy
• ▪Carsten Jung of the IPPR think tank stated on September 28, 2026, that artificial intelligence will have a larger impact on the labour market than the industrial revolution.
• ▪Carsten Jung of the IPPR think tank stated on September 28, 2026, that about 11% of UK jobs, led by secretarial and administrative occupations, are highly exposed to replacement by AI chatbots.
• ▪Carsten Jung of the IPPR think tank warned on September 28, 2026, that a worst-case scenario could see about 8 million UK jobs negatively affected by artificial intelligence.

Anthropic researchers on AI existential risk

• ▪Former Anthropic researcher Jacob Coxon warned in a post on X that AI builders earnestly believe artificial intelligence could kill humanity by the end of the decade.
• ▪Anthropic scientist Evan Hubinger agreed with Jacob Coxon's warning, placing the existential risk of artificial intelligence above 10%.

US stance on AI

• ▪US President Donald Trump has stated that talk of artificial intelligence risks is a hoax.
• ▪US President Donald Trump has warned that controlling the pace of US frontier AI lab research would hand China a significant advantage.

Debatable claims

• ▪Competing national interests make a global AI safety agreement unachievable
• ▪Governments should regulate AI safety rather than relying on developer self-policing
• ▪AI will cause unprecedented job losses that the labor market cannot absorb
• ▪Existential risk from superintelligence is more urgent than current AI misuse