UK funds £100m AI biosecurity lab with OpenAI and Anthropic - The Next Web

The Cabinet Office chose Saturday to announce that the UK will put £100m into a new AI Biosecurity Lab — Saturday being the traditional slot for news you would rather the Sunday papers buried. OpenAI and Anthropic will add up to £200m between them, which is a generous way of saying that the two American labs whose models would presumably be doing the spotting are also, in part, paying for the spotting. Whitehall will call this a partnership. In plainer terms, the vendor now sits inside the firewall.

The lab's opening brief is to find new biological threats early and speed up work on vaccines and treatments, a sentence that will give anyone who lived through 2020 a small, specific flutter. The logic runs roughly like this: a model capable of designing something unpleasant can be met by a larger model reading sequences faster, and the difference between the two is essentially a purchase order. £100m of Treasury money goes in, £200m of San Francisco money comes alongside, and the state supplies the letterhead while the capability arrives from two companies it spent last year promising to get a grip on.

All of it sits inside a £600m AI security package "set out before the Budget", a phrase that translates as: the numbers exist in a press release and will be reconciled with reality at a convenient future date. Britain has spent a fortnight quietly seeking powers over foreign AI firms; it has now resolved the awkwardness by paying them, which is at least a coherent business model. The threats will be found early, the vaccines will arrive faster, and the two firms best placed to cause precisely the sort of thing the lab is looking for will be in the funding round, on the advisory board, and nowhere near a liability clause.