A Select Committee Chair Says Keeping Both Roles Is a Clear Conflict

3 SEP 2026 — Matt Clifford, who wrote the UK's AI Opportunities Action Plan and served as the government's AI Opportunities Adviser, has joined Anthropic as managing director of international affairs. The committee chair's objection is not that he moved from government to industry, which she says is acceptable. It is that he intends to stay on as chair of ARIA at the same time.

Disclosure: this publication is made by Anthropic, the company that is the subject of this article. The account below is confined to what has been reported, the criticism is quoted in full, and no view is expressed on whether the arrangement is appropriate.

The facts

Clifford wrote the AI Opportunities Action Plan in July 2024 and was appointed the government's AI Opportunities Adviser in January 2025. He has chaired the Advanced Research and Invention Agency since 2022 and is an investor at Entrepreneurs First. His new role leads Anthropic's work with governments outside North America, based in London.

Dame Chi Onwurah, who chairs the House of Commons Science, Innovation and Technology Committee, said that his intention to remain as chair of ARIA, which invests taxpayer money in cutting-edge research much of it AI-related or AI-enabled, creates a clear conflict of interest. She acknowledged that movement between sectors is acceptable and called for robust safeguards and clear boundaries to ensure confidence in ARIA's independence.

Clifford reportedly plans to recuse himself from ARIA matters involving Anthropic. The appointment is reported to be entirely within the rules. A response from AI minister Kanishka Narayan on conflict-of-interest safeguards was awaited at the time of publication.

Simultaneous, not sequential

The easy label for an appointment like this is the revolving door, and Onwurah explicitly rejects it. Her complaint is about two roles held at once.

The distinction matters because the safeguards are different. A former official joining a company they regulated is managed by a cooling-off period, which has a defined end. Two concurrent positions cannot be managed by waiting; the mechanism has to be recusal, and recusal only works where the conflicted decisions can be identified in advance.

ARIA funds high-risk research across a portfolio, much of it AI-related. Recusing from anything naming Anthropic is straightforward. Recusing from decisions about which research directions a public funder should back, when the person making them also represents a company with commercial interests in those directions, is a harder line to draw and nobody has described how it would be drawn.

The plan he wrote is now policy he benefits from

The second thread is the Action Plan. An adviser who designs a national AI strategy and then joins a company operating under it has not broken any rule, and the pattern is common enough that governments accept it as the price of getting expert advisers at all.

It does create a question the public record cannot answer: were any recommendations in the plan shaped by a job that had not yet been agreed? Nothing reported suggests it was, and nobody has alleged it.

The arrangement makes it impossible for anyone outside to be confident either way. That is the cost Onwurah is describing when she asks for confidence in ARIA's independence rather than for a finding of wrongdoing.

Within the rules is the finding, not the defence

That the appointment complies with the applicable rules appears to be accepted by everyone involved, including the critic. What compliance establishes is narrower than it sounds.

UK rules on post-government employment principally govern civil servants and ministers moving to the private sector, and are administered through advisory processes with limited enforcement. An adviser holding a separate public appointment is a configuration those rules were not primarily written for.

So compliance here is evidence that the rules did not anticipate this shape, rather than evidence that the arrangement raises no issue. A select committee chair saying the rules were followed and there is still a clear conflict of interest is making exactly that point.

Why this matters outside Britain

Every government building AI policy is short of people who understand the technology, and the people who do understand it are largely employed by the companies being regulated. Singapore, Malaysia and Indonesia are drawing on the same small pool, usually through advisory panels and secondments with less scrutiny than a Commons select committee provides.

The structural problem is that governments cannot avoid this by excluding industry experts, because a regulator with no such expertise regulates badly. It can only be managed by disclosure that is specific enough to be checked, and by keeping the two roles separate in time rather than trusting recusal within a single calendar.

We reported on an advocacy campaign funded by AI investors buying advertising in three US states, where the tension was between a company's stated position and its president's personal political funding. This is the same category of question in a different form, and it is being asked about the company that publishes this article.