Britain launches a £100 million AI biosecurity lab

The British government announced the creation of the "AI Biosecurity Lab" (AIBL) with £100 million in government funding, in addition to commitments from OpenAI and Anthropic reaching up to £200 million. The government stated that the lab will be the world's first organization specialized in AI biosecurity.

The lab aims to develop technical capabilities to help Britain understand, prevent, detect, and respond to biological threats by leveraging advanced artificial intelligence. It will work with national biosecurity centers and the life sciences sector to improve early threat detection and accelerate the discovery of vaccines, treatments, and medical countermeasures.

According to the government, OpenAI will provide £100 million in compute credits, models, and trusted access to its most capable models through the Rosalind Biodefense program. Anthropic will provide up to £100 million in model credits, trusted access, and dedicated technical support. Basecamp Research will also contribute its models and capabilities in biological AI, while scientists and experts from Google DeepMind will collaborate on joint research.

The lab will operate as an independent non-profit organization, partnering with the UK Health Security Agency, the Animal and Plant Health Agency, and the Defence Science and Technology Laboratory, as part of a national AI biosecurity initiative. The government says the lab will also be funded by the public and philanthropic sectors to attract additional investment.

What do these terms mean?

Biosecurity: Measures that reduce the risks of natural or deliberate biological threats, such as infectious diseases or hazardous materials, including prevention, detection, and response.

Compute credits: Financial credits allocated to use processing power or cloud services, usually provided by companies to support research without direct cash transfers.

Frontier models: AI models at the highest levels of capability that require additional safety controls because their misuse could be more dangerous than standard models.