A disclosure first: this site uses Claude, which is made by Anthropic, and Anthropic is one of the partners in the lab described below. We have tried to report only what the government says, and we say where we cannot check it.
On 24 October the Cabinet Office announced the AI Biosecurity Lab, which it calls the first body in the world dedicated to biosecurity with AI. The government is putting in £100 million. OpenAI and Anthropic are offering credits and access, not cash.
What the government says
According to the announcement, the lab sits inside the National AI Biosecurity Initiative and will be an independent, not-for-profit organisation with a scientific committee. It is meant to work with the UK Health Security Agency, the Animal and Plant Health Agency, the Defence Science and Technology Laboratory and the Medicines and Healthcare products Regulatory Agency. Its first aims are the early detection of biological threats and speeding up vaccines and treatments.
Money and credits are not the same
| Contributor | What the government says it gives | What it is |
|---|---|---|
| UK government | £100 million | Public money |
| OpenAI | £100 million of compute and model credits, plus trusted access to models through its Rosalind Biodefense programme | Credits and access |
| Anthropic | Up to £100 million of model credits, trusted access and technical support | Credits and access |
| Google DeepMind and Basecamp Research | Collaboration | No value given |
The government describes a total of up to £300 million: £100 million public and up to £200 million from the companies. That is the government’s own total. The company part is credits to use the companies’ products, so it should not be read as £300 million of spending.
What we do not know
The announcement does not name a director or a location, and gives no start date or timetable for spending the money. It does not say what Google DeepMind will contribute. Nor does it explain what “trusted access” to models means in practice, or how the Rosalind Biodefense programme works.
The Next Web, which cites the announcement, says the lab comes as part of a larger package, with more than £600 million for AI security, including £120 million for the AI Security Institute. We have not checked that figure against a government document. Other reports give different totals, so we use only the numbers in the government’s announcement.
What this means
Our reading: the new body is a public commitment and a set of company promises, and the details that would show how it will work are still to come. Because two of the partners sell the AI models the lab will use, it will matter who decides what the lab does, and how it reports. The announcement says it will be independent and have a scientific committee, but it does not yet say who sits on it.
For another story about Anthropic, see Anthropic’s plan to put 5% of its shares into a public benefit fund. For a different side of AI and safety, see India’s claim that AI agents reached live systems without authorisation.
What we could not confirm
We did not find comment from biosecurity experts outside the government and the companies. We could not confirm the director, location, start date, the value of DeepMind’s contribution or how the credits will be used. All figures are the government’s.
Leave a Reply