He Gives the Money Away. What He Keeps Is the Proximity.

Two quiet letters appeared on gov.uk on the 9th of October 2025. A year on, they still read like a template for how the door turns.

Key · companies amber, places red

The letters come from the Advisory Committee on Business Appointments, the watchdog that vets former ministers' jobs. They clear the Rt Hon Rishi Sunak MP, still the member for Richmond and Northallerton, to advise Microsoft and Anthropic. A third role, at Goldman Sachs, was approved earlier.

The interesting part is not the approval. It is the paragraph beneath it.

The concern, in the watchdog's own words

For Microsoft, Acoba wrote of "a reasonable concern that you could be seen to offer unfair access to, and influence in, the UK government". For Anthropic, "a reasonable concern that your appointment could be seen to offer unfair access and influence". Then the hedge that does the most work: the Committee's "advice is not an endorsement of the appointment". It approved the door and named the risk in the same paragraph.

"Advice is not an endorsement." The watchdog cleared the appointment and named the danger in one letter.

It matters because Britain's claim to lead on AI safety was built by this man. In November 2023 he convened the first global AI Safety Summit at Bletchley Park and launched the AI Safety Institute that other nations copied. At that summit he announced a £2.5bn Microsoft investment in UK datacentres and training. The country sold itself as AI's conscience. Its chief author now sits on the payroll of two of the companies most invested in what that conscience becomes.

The stakes are not abstract. Microsoft holds a five-year agreement with the Crown Commercial Service and, in September 2025, pledged £22bn of UK AI infrastructure. Anthropic signed a government memorandum in February 2025 and still researches with the UK AI Security Institute. Both lobby, in Washington and Brussels, on the rules that govern them.

What is legal, and what is uncomfortable

The conditions work as designed. For two years the roles are ring-fenced to macro-economic and geopolitical strategy: no lobbying, no UK policy advice, no contact with officials, no privileged information. All disclosed, all lawful.

This is not a cash grab. I used the two Acoba letters and the register: Sunak donates every pound, from Microsoft, Anthropic and Goldman, to The Richmond Project, the numeracy charity he founded with his wife, and the register marks it as donated. There is no salary to attack.

That is what makes it a story rather than a scandal. In the AI age the scarce commodity is not capital; it is the appearance of access to the people who write the rules. Acoba named exactly that fear.

The door is not new. Sunak's former adviser, Liam Booth-Smith, is also on Anthropic's payroll, and Nick Clegg ran global affairs at Meta. What is new is the receipt: the watchdog approving the walk-through and warning about it on one page.

Nothing here is illegal. The roles are approved, with conditions, and the money goes to charity. But the guardrails and the guided now turn through the same small door, and a donation cannot buy back the thing that matters: standing where the rules are made. If your firm wants an honest read of who is really shaping it, email brandon@kreostudio.co.uk.