The co-founder of Anthropic, the artificial intelligence company behind the Claude chatbot, has issued a stark warning that AI must not be allowed to develop independently of human oversight. Jack Clark, who also serves as the organisation’s policy chief, argued that the technology risks becoming an uncontrollable force if regulators fail to embed human values into its core. His remarks come as the United Kingdom positions itself at the forefront of a global push to create binding rules for AI, challenging the laissez-faire attitude that has dominated Silicon Valley.
Speaking at a technology summit in London, Clark said that the current trajectory of AI research resembles a ‘race to the bottom’ where companies prioritise capability over safety. He called for mandatory safety testing, transparency requirements, and a new model of ‘human-in-the-loop’ governance that ensures people remain the ultimate decision-makers. ‘We have to build systems that understand our intent, our ethics and our limitations. If we don't, we risk creating digital entities that operate with values that are at best indifferent and at worst hostile to humanity,’ he said.
The urgency of Clark’s message was underscored by recent breakthroughs in large language models and generative AI that have stunned even their creators. Many experts worry that the next generation of systems will be ‘beyond the comprehension of any single human’, making oversight impossible. The UK government, which hosted the world's first global AI safety summit at Bletchley Park last year, has now proposed legislation that would classify high-risk AI systems and impose fines of up to 10 per cent of global turnover for non-compliance. The move has attracted praise from civil society groups but criticism from some tech companies who fear it will stifle innovation.
‘The UK is demonstrating rare leadership in a domain that desperately needs it,’ Clark added. ‘But no country can solve this alone. We need a coalition of governments, researchers and citizens to design a system of checks and balances before it's too late.’ His comments align with growing calls from Europe and North America for an international treaty on artificial intelligence, similar to those governing nuclear weapons or climate change.
Critics, however, argue that the UK’s push risks creating a fragmented regulatory landscape where companies simply relocate to jurisdictions with weaker rules. They point to the rapid deployment of generative AI tools in consumer products as evidence that the market has already outpaced policymakers. Clark acknowledged these concerns but insisted that the alternative is far worse. ‘Inaction is a decision. It is the decision to let a handful of corporations shape our collective future without democratic oversight,’ he said.
For the average person, the debate may seem abstract, but the implications are tangible. From algorithmic bias in job applications to AI-powered disinformation in elections, systems are already making decisions that affect millions. The Anthropic co-founder’s warning is a reminder that the technology we build today will define the boundaries of human agency tomorrow. Whether the UK’s regulatory push will create a template for the world or become a cautionary tale about the limits of governance depends on how quickly other nations follow suit.
As the sun set over the Thames, Clark left the stage with a final thought: ‘We have the opportunity to design the future we want. But we are running out of time to choose.’
