The co-founder of Anthropic, the AI safety company behind the Claude model, has issued a stark warning to HM Treasury: Britain's failure to build sovereign oversight of artificial intelligence is a direct threat to national security. Speaking at a closed-door briefing in Whitehall, the executive argued that the UK's current reliance on foreign AI infrastructure and governance frameworks leaves it dangerously exposed to economic sabotage, cyber warfare, and loss of democratic control.
'We are sleepwalking into a digital colony', the co-founder is reported to have said. 'If Britain does not develop its own AI auditing capabilities, ethical standards, and regulatory muscle, it will be forced to accept the values of other nations embedded in the algorithms that run its public services.' The warning comes as the Treasury accelerates its adoption of AI for tax collection, benefits assessment, and fraud detection, raising fears of algorithmic bias and data sovereignty breaches.
The Anthropic executive specifically called for a British AI Safety Institute with teeth, not just advisory powers. He argued for mandatory stress testing of all AI models used in critical infrastructure, akin to the Bank of England's stress tests for financial institutions. 'We need a digital version of the Bank of England's oversight', he said. 'AI models can crash economies faster than any subprime mortgage crisis.'
HM Treasury has faced mounting pressure from tech ethics groups and MPs on the Joint Committee on the National Security Strategy to slow down AI procurement. Critics point to the 2023 Horizon scandal as a cautionary tale of what happens when automated systems lack proper oversight. 'The Post Office's Horizon software ruined thousands of lives because there was no independent British oversight of its algorithms', said Labour MP for Tech Policy. 'We cannot repeat that mistake at scale.'
The warning also highlighted the geopolitical dimension. The US and China are investing billions into national AI projects, while the EU is pushing through the AI Act with extraterritorial implications. Britain's post-Brexit desire for agile regulation may leave it as a 'rule-taker rather than rule-maker', the Anthropic co-founder argued. 'If we don't build our own AI standards, we will be playing by rules written in Beijing or Silicon Valley.'
The Treasury responded by pointing to its recent £100 million investment in the AI Safety Institute and the formation of the Frontier AI Taskforce. However, insiders say these bodies lack enforcement powers and are heavily reliant on US cloud providers for compute. 'We are outsourcing the brain of our government to Amazon Web Services and Microsoft Azure', a former GCHQ analyst told the Guardian. 'That is a national security risk of the highest order.'
The Anthropic co-founder's intervention is seen as significant because his company is often considered more aligned with British values of rigorous safety testing than its US counterparts. He urged the Treasury to mandate that any AI system used by the British government must have its model cards and bias audits made publicly available. 'Transparency is the only antidote to algorithmic authoritarianism', he said.
Industry reaction has been mixed. Some tech leaders welcome the call for oversight as a way to build public trust. Others warn that heavy regulation could stifle innovation. 'We need a British way that is neither the Wild West of America nor the bureaucratic forest of Europe', said a spokesperson for TechUK.
As the Treasury consults on its AI regulation white paper, the Anthropic warning adds a new urgency. The question remains whether Britain can assert its digital sovereignty before the algorithms that run its hospitals, courts, and army become black boxes controlled from abroad.
