Three thousand delegates at the London Tech Summit heard an extraordinary plea this morning from Jack Clark, co-founder of the artificial intelligence company Anthropic. Speaking from a stage bathed in blue light, Clark delivered a stark warning: the next generation of AI systems must not be allowed to develop without robust human oversight.
Clark, whose firm is behind the Claude language model, argued that the race to build ever more powerful AI has become a ‘lumbering giant’ without a steering wheel. “We are building machines that can write code, manipulate images and reason at graduate level,” he said. “But we have not yet built the equivalent of a digital air traffic control system for them. That must change before it is too late.”
The summit’s organisers immediately seized on the moment, unveiling a draft for a UK-led International Charter on AI Governance. The document, which is expected to be floated at the United Nations General Assembly next month, would commit signatories to three core principles: transparency of training data, mandatory human override for high-risk decisions, and independent safety audits for any system capable of causing significant harm.
Clark’s intervention comes amid growing unease in the technology community about the pace of deployment. Just last week, a group of prominent engineers published an open letter warning that frontier models could soon be capable of autonomous cyberattacks or the creation of biological weapons. “The technology is not sentient, and it may never be,” Clark told the audience. “But it is already powerful enough to amplify human error on a catastrophic scale. The only safeguard is a system that answers to elected governments, not to corporate balance sheets.”
The proposed charter has already attracted support from several unlikely quarters. The German delegation, which initially resisted any form of international regulation, has now signalled its willingness to participate. The British government has allocated £100 million for an AI Safety Institute that would serve as the charter’s technical backbone. And a coalition of Commonwealth nations has agreed to sponsor the text at the UN.
Critics, however, argue that the charter is toothless. “This is a beautifully written piece of moral theatre,” said Dr. Helena Cross, a digital ethics researcher at Oxford. “But it has no enforcement mechanism. What happens if a company in Singapore or a state in Beijing decides to ignore it? The charter will be a piece of paper, and the real power will remain with those who hold the data and the compute.”
Clark acknowledges the problem but insists that something is better than nothing. “Every major technological shift has required a framework of rules,” he said. “Nuclear energy has the IAEA. Aviation has ICAO. It is time for AI to have its own global body. Yes, it will start weak. But it can grow stronger as nations realise that chaos is no one’s friend.”
The summit continues tomorrow with a closed-door session on what the charter should say about military AI. The stakes could not be higher. As Clark put it in his closing remarks: “We are writing the rules for a power that could reshape every aspect of human life. If we get this wrong, history will not forgive us.”
For now, the world watches London. The real question is whether this city, once the capital of an industrial revolution, can lead the way in managing the digital one.
