Sunak takes Microsoft and Anthropic adviser roles under watchdog curbs

Former UK prime minister Rishi Sunak will advise both Microsoft and Anthropic while remaining a backbench MP—posts cleared with a two-year lobbying ban and strict limits on UK policy work, according to ACoBA’s formal advice letter. The timing lands as the Bank of England warns AI-driven valuations look stretched, raising the temperature around Big Tech’s political ties and market risk.

Sunak’s remit at both companies is pitched as high-level strategy on macroeconomic and geopolitical trends. Microsoft also gets him on stage at its annual summit. Anthropic gets counsel from the host of the UK’s 2023 AI Safety Summit. Both firms say the roles are “internally focused” and ringfenced from lobbying. Sunak says he’ll donate the pay to The Richmond Project, the numeracy charity he founded with his wife, Akshata Murty.

What’s actually new

This isn’t one politician taking one post. It’s AI companies formalizing a political bench—hiring people who understand how rules get written and how they are enforced. Anthropic has already brought on Biden-era policy hands such as Tarun Chhabra and Elizabeth Kelly. Sunak extends that approach internationally: a Conservative leader with firsthand experience launching the AI Safety Institute and convening governments and labs. That network travels well. So does the playbook.

The watchdog conditions are boilerplate but notable. Sunak can’t lobby UK officials for two years from leaving office, can’t draw on privileged information, and can’t advise on UK-specific policy. The committee also flagged his one-to-one meetings with Bill Gates while in No. 10 and his role in shepherding Microsoft’s UK investment announcements. Anthropic, it noted, was already a government stakeholder on public-service AI.

The ringfencing problem

“Internally focused” sounds clean on paper. Reality is messier.

Sunak’s edge is not a phone call to Whitehall. It’s judgment—what arguments land in Brussels, which trade-off regulators will accept, where compliance can become a moat, and how sovereign AI plans reshape procurement. That knowledge doesn’t require outreach to be valuable. It shapes roadmaps, partnerships, and messaging. Firewalls block emails, not intuition.

For Anthropic, the hire underwrites its safety-first brand. The company sells governance as part of the product, and senior political advisers signal exactly that to risk-averse buyers. It’s credibility by résumé. For Microsoft, the prize is wider. The company straddles consumer AI, enterprise software, and a hyperscale cloud exposed to diverging national rules. A seasoned political operator helps navigate that patchwork. No lobbying needed.

Advising rivals, in practice

On paper, the conflicts are manageable. Microsoft backs OpenAI; Anthropic is its fiercest enterprise rival; both run on clouds that include Azure. These roles are scoped to geopolitics and macro trends, not model choices or deal strategy. That is the theory.

In practice, context leaks. Understanding one lab’s constraints—compute, capital, compliance—inevitably informs how you frame risks and timelines at another. The separation exists in job descriptions, not in cognitive compartments. It’s a line to watch.

Valuations, risk, and timing

The backdrop is frothy. On October 8, the Bank of England’s Financial Policy Committee warned that AI-linked equity valuations “appear stretched,” and that a sharp correction would hit the broader economy. Concentration risk is high. Expectations are higher. One disappointment can ripple across the complex.

Sunak’s appointments landed days later. Anthropic’s latest private-market valuation has soared—The Times puts it around $183 billion. Microsoft’s market cap bakes in substantial AI optionality. None of this implies imminent trouble. It does mean shareholders and regulators will scrutinize how companies wield political access while markets price perfection. Volatility will test the narrative.

The new post-office pipeline

Tech has long recruited ex-officials. The pace and purpose feel different now. AI companies are assembling policy shops in-house, competing for the same ex-ministers and national-security aides they’ll soon brief across the table. It’s a structural response to regulatory heat and a bet that institutional memory is a moat.

Sunak’s trajectory—Goldman Sachs adviser in July, now dual AI roles—fits the mold. The novelty is the speed with which AI labs and platforms have professionalized their political operations. They are hiring like think tanks.

Three tests ahead

First, enforcement. ACoBA’s rules matter only if companies treat them as hard constraints when outcomes get tight. Second, market discipline. If AI’s premium deflates, the appetite for marquee advisers may cool quickly. Third, rulemaking. Governments could tighten the revolving-door code as AI lobbying intensifies. Those choices will set the norm for who can advise whom, and on what.

Sunak’s dual posts are an early case study. They probe how far “ringfenced” can stretch, whether political capital compounds across rivals, and how fast ethics rules written for an earlier tech era can adapt. That’s the real story.

Why this matters

• AI firms are building policy muscle at speed, turning institutional know-how into a competitive asset as rules, procurement, and subsidies shape winners.
• Ringfencing curbs direct lobbying, but not judgment; political insight can still influence product, partnerships, and go-to-market in ways that matter.