← Back to feed

Matt Clifford built the UK’s AI policy infrastructure. He co-founded Entrepreneur First, advised the government on frontier AI strategy, and in 2022 became the founding chair of the Advanced Research and Invention Agency — the publicly funded body designed to back high-risk, high-reward research including AI. On September 2, Anthropic announced he would join as managing director of international affairs, the role responsible for the lab’s government relations outside North America. Five days later, he announced his resignation from ARIA.

The sequence was compressed but the pressure was immediate. Chi Onwurah, chair of the House of Commons Science, Innovation and Technology Committee, called the dual-role arrangement a “clear conflict of interest” and wrote to the government seeking answers on how it arose. ARIA funds AI research. Anthropic is an AI lab. Clifford, as ARIA chair, influenced which research received government backing. As Anthropic’s international affairs lead, he would represent the lab’s interests to the same governments ARIA operates within.

The timeline

• September 2, 2026: Anthropic announces Clifford as MD of International Affairs
• Pre-September 7: Onwurah raises conflict of interest publicly via the Science and Technology Committee
• September 7, 2026: Clifford announces resignation from ARIA on X: “I have decided to step down to ensure my new role at Anthropic doesn’t become a distraction from Aria’s incredible work.”
• November 6, 2026: Clifford’s scheduled last day at ARIA, at the request of business and science secretary Jonathan Reynolds to allow time for a replacement search

The departure was described by Clifford as voluntary. Onwurah called it the “right” decision while maintaining that “important questions remain.” She has requested from the government a full account of how the appointment was cleared, what conflict-of-interest assessment was conducted, and what safeguards are in place during the seven-week transition period.

What this reveals

Anthropic’s international affairs role is not a token government relations post. The lab’s previous international moves — including a string of government MOU signings across Europe and Asia in 2025 — suggest the role is substantive, intended to shape how regulators treat frontier AI models and how public procurement of AI services is structured.

Hiring the architect of a government research agency to run those conversations is a deliberately close orbit. Clifford’s relationships and institutional knowledge from ARIA represent exactly the kind of embedded access that lobbying cannot easily replicate.

The UK’s position makes this more pointed. ARIA operates under the Department for Science, Innovation and Technology. It funds research the government considers strategic. A chair with detailed knowledge of ARIA’s priorities, funding pipeline, and internal deliberations joining a company that could receive or be affected by that funding crosses a line that a recusal arrangement — the initial proposal — could not adequately address.

The broader pattern

The Clifford episode fits a pattern visible across major AI labs in 2026. OpenAI hired former national security officials. Google DeepMind placed researchers in advisory roles to multiple governments simultaneously. The labs are competing not just on benchmark performance but on regulatory positioning — trying to ensure that the rules being written about frontier AI are written with their input, or at minimum, not against their interests.

Hiring from within government structures accelerates that positioning. The conflict of interest concern at ARIA is the most direct instance to date of a government drawing a line and asking for distance. Whether the underlying arrangement — former government architect now running AI lab government relations — changes in practice, rather than just in title, remains an open question that Onwurah has made clear she intends to pursue.