On Saturday the UK government announced the AI Biosecurity Lab, which it calls the world's first organization dedicated to AI and biosecurity. The government is putting in £100 million. OpenAI is committing £100 million and Anthropic up to £100 million, mostly in model credits and early access to their strongest models, and Google DeepMind and the London company Basecamp Research will work with it on research. The lab will be an independent nonprofit and will start with two jobs: spotting emerging biological threats earlier, and speeding up vaccines and treatments.

The government's own reasoning is blunt. AI can speed up disease research, it says, but more capable systems "also create new biological risks and could lower barriers to harmful activity." So the lab will work alongside the UK Health Security Agency, the Animal and Plant Health Agency and the Defence Science and Technology Laboratory, and with the medicines regulator to help shape how AI-designed medicines are regulated. It is part of a £165 million National AI Biosecurity Initiative that the Chancellor, John Healey, is folding into next week's Budget. The companies whose systems raise the risk are now paying to help track it. (UK government, AI Biosecurity Lab)

The lab is one piece of a larger UK package worth more than £600 million, according to the tech industry body techUK. The AI Security Institute, the government team that tests new models before and after release, receives another £120 million. A £60 million National Cyber Shield will use AI agents to defend British networks, including a £25 million fund for UK cyber firms and early work on a regulatory sandbox for agentic cyber defence in critical infrastructure, the public sector and industry.

A new £70 million National Exposure and Real-Time Vulnerability Centre will use AI to find weak spots across critical infrastructure and the wider economy. Ciaran Martin, the former head of the National Cyber Security Centre, said the week had been "a stark reminder" that AI exposes digital weaknesses and "brings new ways of exploiting them." If your company runs anything Britain calls critical infrastructure, these programs are aimed squarely at you. (HM Treasury)

The worry behind all this is no longer theoretical. The Arms Control Association's October journal walks through work by Stanford University and the Arc Institute, published in Science in August, in which genomic language models designed whole virus genomes not found in nature. The scientists generated thousands of designs, chemically built 285, and ended up with 16 viruses that worked. They were bacteriophages, which infect bacteria and pose no risk to people, and the aim was new treatments for drug-resistant infections.

Biosecurity experts noticed the other use. Moritz Hanke of the Johns Hopkins Center for Health Security told The New York Times someone could ask such a model for a flu genome made more transmissible or more lethal. In September Anthropic said it had caught foreign users asking Claude for information useful for making biological weapons, and cut them off. AI makers now have to police the same capability they sell to drug developers. Britain's lab is one answer. (Arms Control Association)

Washington is drawing lines around the same technologies with money. On Thursday the US Treasury proposed rules under the COINS Act, passed last December, that would require American investors to notify it of certain deals with companies tied to China and other countries of concern, and ban others outright. The sectors are chips, quantum, supercomputing, hypersonics, AI, and a new addition: biotechnology, including tools for biological design, bioinformatics and automated scientific discovery.

Following an executive order from September, the proposal swaps "AI systems" for "super intelligence systems" throughout. Deals worth $50,000 or less over 12 months are spared from notifying Treasury, but not from the bans. Notices are due within 30 days of closing, and Treasury could force a sale of a banned stake. Comments close December 7. If your company or its venture arm invests abroad, the deal team should read this now. (US Treasury; fact sheet)

Already filed, if you want the run-up: The Clerk (a US appeals court taking a case away from a judge over an AI-drafted order), The Tab (Europe's Parliament fighting over who pays when AI does harm), and The Footer (Brazil's top court fining lawyers for hiding orders to its AI).

Also on the shelf: The Loop (Japan's rulebook for 180,000 civil servants using AI agents), The Stake (Anthropic putting 5% of its shares into a public benefit fund) and The Brake (Satya Nadella on treating models as insider risks). Read today beside The Stake. Last week Anthropic put part of its value into a public fund. This week two AI labs are paying a government to watch for misuse of what they build.

Scientists have been here before. In 1974, worried that new gene-splicing techniques could release dangerous organisms, a committee of leading researchers called on scientists to pause recombinant DNA experiments. Paul Berg, the Stanford biochemist who had pioneered the method, had already held back from the riskiest step of his own experiment after colleagues raised alarms.

In February 1975 Berg, Maxine Singer and others gathered about 140 people at the Asilomar conference center on the California coast and drew up voluntary rules: containment matched to the estimated risk, weakened bacteria that couldn't survive outside the lab, and training for everyone involved. Research resumed under those rules. The field policed itself before anyone made it. Fifty years on, AI labs are being asked to do the same for a far faster technology. (Asilomar Conference on Recombinant DNA)

Put these to management. Do any of our AI tools, or our vendors' tools, touch biology, chemistry or lab work: drug discovery, research, diagnostics, even supply chains for lab equipment? If someone tried to misuse them, would we see it, and who would we tell?

Then look at the money. Does our company or its venture arm hold stakes in AI, chip or biotech firms with ties to China, and would Treasury's new rules catch them? Have we costed what it would take to unwind one if the answer is yes?

The Move

Britain just told the world where it thinks the sharpest AI risk sits: in biology. Two of the biggest model makers agreed and put £200 million in credits behind it. Rules usually follow where the money and the labs go.

Ask the chief risk officer for one page before the next board meeting: every place our AI touches biology or lab science, how we would spot misuse, and any investment that Treasury's new rules could reach.

The lab is open. Know what yours is doing.

Seven bullets. One move.