2 min read AI-generated

Jack Clark at Oxford: 'AI Could Kill Us All' - and Anthropic Is Selling It Anyway

Copy article as Markdown

Anthropic's co-founder warns of existential AI risks at Oxford. Meanwhile, his company markets Claude Code more aggressively than ever. Time calls it 'A Tale of Two Anthropics'.

Featured image for "Jack Clark at Oxford: 'AI Could Kill Us All' - and Anthropic Is Selling It Anyway"

On May 22, Jack Clark — Anthropic co-founder and head of the newly established Anthropic Institute — delivered the second Cosmos Lecture at the University of Oxford. His message wasn’t exactly reassuring: AI poses a ‘non-zero’ chance of killing everyone on the planet. By 2028, perhaps sooner, AI systems could achieve recursive self-improvement — a point after which predictions become very difficult.

The Two Faces of Anthropic

Time Magazine used the week’s events in London and Oxford to publish a profile called ‘A Tale of Two Anthropics’. Because while Clark was warning about existential risks at Oxford, Anthropic had just hosted its Code with Claude event in London a few days earlier — with hundreds of developers enthusiastically exploring the latest Claude Code features.

This tension isn’t accidental. Anthropic makes money from a product that its own founders say could endanger humanity. Clark doesn’t try to hide it. In his speech, he compared the situation to early nuclear research: you can acknowledge the risks while still working to make the technology safe.

What Clark Actually Said

Beyond the existential warning, Clark also made more optimistic predictions: AI will help win a Nobel Prize within twelve months. He forecasted bipedal trade robots, AI-run companies, and massive societal changes.

On the topic of Mythos — Anthropic’s most powerful model, which just found over 10,000 security vulnerabilities — Clark said they deliberately chose not to release it because the risks were too great. Instead, they gave controlled access to roughly 50 partners.

My Take

Anthropic’s duality is both fascinating and refreshingly honest. No other AI company talks this openly about the risks of its own technology — while simultaneously selling it this successfully. Whether this balancing act works long-term depends on whether Anthropic can actually pull off the tightrope walk between growth and responsibility. Clark’s Oxford speech shows the company isn’t ignoring the tension — it’s actively confronting it. Whether that’s enough is a different question entirely.

Sources: