AI poses risks to national security, Burnham says after Anthropic expert says tech ‘could kill all humans’
Andy Burnham has acknowledged that artificial intelligence poses risks to UK national security, responding to Commons questions after an Anthropic safety researcher warned the technology could pose an existential threat to humanity. The exchange came after reports that Anthropic had withheld its newest AI model from the UK's AI Safety Institute, prompting concern from MPs about whether Britain has adequate oversight of frontier AI development being carried out by private companies.
Anthropic researcher Evan Hubinger said he believed there was a greater than 10 per cent chance of AI causing a species-ending event within a decade, adding the firm has "no plan to solve alignment for superintelligence". His comments followed the resignation of fellow Anthropic researcher Jacob Coxon, who accused AI labs including Anthropic and OpenAI of "gambling with our lives" through competitive pressure to build self-improving systems. Former minister Darren Jones has called for a new multi-national treaty to regulate superintelligence development, while the Cabinet Office insisted the UK's AI Security Institute continues to work closely with industry, including having tested OpenAI's GPT-6 Astra before release.
In short
- Burnham admits AI poses UK national security risks amid safety warnings
- Anthropic researcher says AI could kill all humans within a decade
- Ex-minister Darren Jones calls for new international AI safety treaty
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Frontier AI could bring serious national security and societal risks, and the warnings from researchers justify stronger independent scrutiny. Requiring companies to share models with safety institutes and negotiating international rules could help governments assess dangers that firms may struggle to manage amid competitive pressure.
The case against
Warnings about catastrophic AI risks remain uncertain, so sweeping restrictions could impede useful research and economic gains without reliably preventing development elsewhere. Close cooperation between companies and safety institutes may offer a more practical way to test systems while preserving innovation and national security capabilities.
Explained simply
Artificial intelligence systems are built by technology companies and can carry out tasks such as writing, answering questions and analysing information. The most advanced systems are often called frontier AI; some researchers are studying whether future versions could become much more capable and difficult to control.
Anthropic is an AI company, and its researchers have raised concerns about the risks of increasingly capable systems. The UK’s AI Security Institute assesses advanced AI and works with companies, while MPs question how the government can oversee development largely taking place in private firms.
The debate matters because AI could affect national security and public safety as its uses expand. Governments are considering how to assess and manage those risks, including whether cooperation between countries may be needed.
AI Art Culture Government Politics Technology