In a move that has sent ripples through the tech world, Anthropic, the AI safety company co-founded by former OpenAI researchers, has suspended several of its tools following a direct intervention from US national security officials. The decision, confirmed by sources close to the company, comes amid fears that its advanced language models could be weaponised for disinformation or cyber attacks. British Technology Secretary Peter Kyle has now demanded a full briefing, warning that 'the era of self-regulation for frontier AI is over'.
The suspended tools include a text-generation system capable of mimicking human writing styles with eerie precision, and a code-writing assistant that security analysts claim could automate the creation of malware. Anthropic’s CEO Dario Amodei stated that the company acted 'out of an abundance of caution' after receiving classified intelligence about potential misuse by state-backed actors. However, critics argue that the company should have foreseen these risks earlier, given its own research on 'sleeper agent' attacks in AI models.
Kyle’s response has been characteristically blunt. In an emergency statement, he declared: 'British citizens deserve to know how their data is being used and what safeguards are in place. We cannot have a patchwork of corporate policies dictating global security. I have requested a detailed timeline of Anthropic’s actions, including the exact nature of the US concerns and what metrics were used to justify the shutdown.' The Minister has also hinted at new legislation that would require all AI developers to register high-risk systems with a UK regulator.
The incident has reignited debates about digital sovereignty. While the US often acts unilaterally in tech matters, the UK and EU have pushed for multilateral frameworks. 'This is a classic example of why we need a global AI watchdog,' said Dr. Priya Nair, a Cambridge researcher specialising in AI ethics. 'When a single government leans on a company, it creates a black box. We need transparency not just for users, but for other nations.'
Anthropic’s share price dropped 8% in after-hours trading as investors digested the implications. The company, which bills itself as the 'safety-first' AI lab, now faces a reputational crisis. Its tools were used by thousands of developers, including those building mental health chatbots and educational software. The sudden halt has left many in limbo, wondering if their projects will ever resume.
On the streets of London, the mood is mixed. Some see the suspension as necessary: 'If it stops fake news or cyber attacks, I’m for it,' said Samira, a small business owner. Others worry about censorship. 'This feels like the start of a slippery slope,' argued James, a software engineer. 'Who decides what’s dangerous? The government can’t even regulate social media properly.'
The technology community is split. Y Combinator, the startup accelerator, has urged caution: 'We support safety but not punitive measures that stifle innovation.' Meanwhile, the Electronic Frontier Foundation warned that 'security concerns can be a trojan horse for surveillance.'
In a blog post published late Tuesday, Anthropic tried to reassure users. 'We are working with US and UK authorities to implement robust safeguards. Our goal is to resume services within 14 days, with enhanced monitoring and explainability features.' But the company offered few details on what those features would entail.
As the sun sets on this crisis, one thing is clear: the honeymoon period for AI companies is over. Governments are no longer content to let Silicon Valley dictate the terms of engagement. For Anthropic, the question is not just when its tools will return, but whether trust can be rebuilt. And for the rest of us, the spectre of a fully transparent AI that is simultaneously secure feels more distant than ever.
This story is developing. We will bring you updates as they happen.
