A judge, not a lobbyist, is now running Anthropic's politics

On Tuesday, August 4, Anthropic posted something to its newsroom that wasn't a model, a benchmark, or a price cut. The headline read "Mariano-Florentino (Tino) Cuéllar to join Anthropic as Chief Global Affairs Officer." It was a hiring notice. And within hours it had been picked up by everyone from The Recorder, California's legal trade paper, to The Harvard Crimson — because of where the hire came from.

Cuéllar sat on the Supreme Court of California from January 2015 to October 2021. That's the highest court of the most populous state in the country, and its rulings routinely pull the rest of American privacy and consumer law along behind them. He resigned that seat voluntarily to become the tenth president of the Carnegie Endowment for International Peace, the Washington institution Andrew Carnegie founded in 1910, which runs scholars in roughly 20 countries. He spent nearly five years there. Now he runs policy for an AI company.

Here's the deal on the job itself. The title is Chief Global Affairs Officer, and it's new — Anthropic had never had one. Per the company's own announcement, he will lead "the company's work on policy, strategic international engagement, and government relationships worldwide." He reports directly to President Daniela Amodei. And there's an unusual procedural wrinkle attached: Cuéllar had been a Trustee of Anthropic's Long-Term Benefit Trust since January 2026, and he stepped down from the Trust in order to join the company. Someone who was overseeing Anthropic is now working for it.

The timing is what makes this more than an org-chart update. Anthropic is currently suing the United States government. In early March the Defense Department designated the company a national security "supply chain risk," which functionally locked it out of federal procurement; four days later Anthropic filed suit in the Northern District of California. In mid-June a Commerce Department export control order forced the company to switch off its two top-tier models for every customer worldwide for about three weeks. Neither episode is fully resolved. That's the moment Anthropic chose to create a C-suite seat for global policy.

Cuéllar's own line in the announcement tells you what kind of hire this is. "Policymakers in the US and around the world are increasingly realizing that we are at a critical inflection point when it comes to how we govern and develop artificial intelligence," he said. "Democracies must set the terms on which this technology advances, and there is no more consequential place to be shaping that work right now than Anthropic." That's not the language of a government-relations operator. That's a constitutional lawyer talking. That's the part that matters.

From Matamoros to the California Supreme Court

Cuéllar's résumé barely overlaps with a typical AI executive's. He was born in July 1972 in Matamoros, Tamaulipas, Mexico, immigrated to the US at 14, and grew up in the border town of Calexico, California. Harvard for his BA in 1993, magna cum laude. Yale Law School for his JD in 1997. A Stanford PhD in political science in 2000. When Governor Jerry Brown nominated him in July 2014 and he was sworn in on January 5, 2015, he became the first Mexican immigrant ever to serve on California's highest court.

His government service spans three presidential administrations. In Bill Clinton's second term he worked at the Treasury Department as Senior Advisor to the Under Secretary for Enforcement, handling international money laundering and firearms trafficking. In Barack Obama's first term, in 2009 and 2010, he was Special Assistant to the President for Justice and Regulatory Policy at the Domestic Policy Council, running the teams on criminal justice, public health, immigration, and regulatory reform. He later served on the President's Intelligence Advisory Board and the State Department's Foreign Affairs Policy Board. For a company whose top models were frozen by an export control order six weeks ago, sitting at intelligence-community and State Department advisory tables is not decorative experience.

His academic career points at the same problem from a different angle. He taught at Stanford for two decades, directing the Freeman Spogli Institute for International Studies, co-directing the Center for International Security and Cooperation, and running the Stanford Cyber Initiative. He is the Cameron Schrier Family Professor at Stanford Law School and a senior fellow at Stanford's Institute for Human-Centered AI. Anthropic's announcement notes he began organizing AI classes nearly a decade ago, and he was appointed to the National Academy of Sciences' Committee on Responsible Computing Research.

The most load-bearing item on the list, though, is California. Governor Gavin Newsom convened a Joint Policy Working Group on AI Frontier Models in September 2024, and Cuéllar co-led it alongside Stanford HAI co-director Fei-Fei Li and UC Berkeley's Jennifer Chayes. Their report, "The California Report on Frontier AI Policy," landed on June 17, 2025, and became the intellectual foundation for SB 53 — the Transparency in Frontier Artificial Intelligence Act — which Newsom signed on September 29, 2025 as the first US statute written specifically for frontier AI. Cuéllar himself said the law reflected the transparency and "trust but verify" principles the report had laid out. So: one of the people who drafted the blueprint for the law that regulates Anthropic now works at Anthropic.

His Carnegie tenure has receipts too. The January 15, 2026 announcement of his departure credits him with launching new programs on climate, global order, and AI, opening offices in Singapore and Palo Alto, and standing up the Carnegie Russia Eurasia Center in Berlin. He also co-chaired a bipartisan Task Force on Nuclear Proliferation and American Security. Board chair Jane Hartley said his "innovative leadership helped Carnegie weather the geopolitical storms and forge new workstreams." His successor, announced June 11, is former Director of National Intelligence Avril Haines, who takes over on September 28.

And here's a twist worth flagging: Anthropic wasn't the plan. That January announcement said Cuéllar would return to Stanford in July to lead the Center for Advanced Study in the Behavioral Sciences and to succeed former Stanford president John Hennessy as director of the Knight-Hennessy Scholars Program. Six months later the destination changed. Per The Crimson, he is taking a leave of absence from Stanford Law School to join Anthropic. He has served on the Harvard Corporation, Harvard's senior governing board, since February 2019, and chairs the William and Flora Hewlett Foundation board.

What the job actually consists of

Anthropic's policy organization has been rebuilt over the past six months. Co-founder Jack Clark, who ran policy from the company's 2021 founding, moved in early 2026 into a new role as Head of Public Benefit and now leads the Anthropic Institute, launched March 11, which folds together the Frontier Red Team, the Societal Impacts team, and the Economic Research team. Day-to-day public policy is led by Sarah Heck as Head of Public Policy, who came from Stripe and, before that, the White House National Security Council. Cuéllar's role sits on top of all of it and adds international and government relations.

Why that layer became necessary is obvious once you line up Anthropic's 2026 the way a docket clerk would.

Date Event Status
July 2025 ~$200M contract via DoD's Chief Digital and AI Office; Claude becomes first frontier AI cleared for classified military use Signed
Jan–Feb 2026 Standoff over Anthropic's ban on autonomous lethal weapons and mass domestic surveillance; DoD demands "any lawful use" terms Talks collapse
Early March 2026 DoD designates Anthropic a national security supply chain risk Locked out of federal procurement
March 9, 2026 Anthropic sues in N.D. Cal. (docket 3:26-cv-01996) Litigation opens
March 26, 2026 Judge Rita F. Lin grants preliminary injunction: "classic illegal First Amendment retaliation" Partial win for Anthropic
April 8, 2026 D.C. appeals court denies emergency motion to lift the designation Loss for Anthropic
Mid-June 2026 Commerce export control order halts foreign-national access; Anthropic disables two top models for all customers Service outage
June 30, 2026 Anthropic confirms export controls lifted Resolved in ~3 weeks
July–Aug 2026 Cross-motions for summary judgment; Lin calls the government's position "really troubling and at odds with the First Amendment" Final ruling pending

Read that table and the story of Anthropic's first half of 2026 is a litigation calendar, not a product roadmap. The June episode is the sharpest illustration. Per NPR and CNBC reporting, the administration issued an order suspending foreign nationals' access to the company's top-tier models, and Anthropic's response was to turn the models off for everyone. Commerce lifted the controls at the end of the month. That is regulatory risk reaching directly into the revenue pipe and switching it off.

The money tells the same story. Federal lobbying disclosures compiled by Axios and CNBC show Anthropic spent $1.6 million in Q1 2026 and $1.97 million in Q2, for $3.53 million in the first half — more in six months than the $3.1 million it spent in all of 2025. OpenAI's first-half total was $2.22 million. On top of that sits political money: Anthropic gave $20 million to Public First Action, a group backing AI regulation, on February 12, then another $20 million on July 22, for $40 million total. The group's stated priorities include opposing federal preemption of state AI rules without a strong federal standard, plus chip export controls and rules on high-risk capabilities like AI-enabled bioweapons.

On the other side sits Leading the Future, launched in January at a reported ~$125 million, with $25 million from OpenAI president Greg Brockman and $25 million from a16z. Its top priority is exactly the thing Public First Action opposes: federal preemption of state AI law. So Anthropic and OpenAI aren't only competing on models — they are funding opposite sides of a fight over whether US states get to regulate AI at all. Bringing in a former appellate justice reads as a decision to argue that fight as a legal and institutional question rather than a lobbying one.

Who gets what out of this

Anthropic gets three things. First, actual qualifications rather than the appearance of them. A First Amendment retaliation suit, an export control review, and a preemption fight are all legal problems. Putting a former justice at the top of policy changes the grammar of everything the company says externally. Cuéllar spent nearly seven years writing majority opinions on technology, privacy, international agreements, and separation of powers. Someone who knows which arguments survive a courtroom is worth considerably more than someone who can write a press release about respecting regulation.

Second, international coverage. Anthropic now has offices in 12 cities. In Asia-Pacific that means Tokyo, Bengaluru, and Seoul; in Europe, London, Dublin, and Zurich, with Paris and Munich added. By the company's own figures, Asia-Pacific run-rate revenue grew more than tenfold in a year and EMEA more than ninefold, with EMEA headcount tripling. At the Tokyo opening Anthropic signed a memorandum of cooperation with the Japan AI Safety Institute, and CEO Dario Amodei met Prime Minister Takaichi. That's well past the point where product leads can handle government relationships as a side task. Running an institution with scholars in 20 countries is directly relevant experience.

Third, legitimacy — and this is the double-edged one. Anthropic's public position is that it's the frontier lab that welcomes regulation, and that position always draws the same objection: that supporting regulation is a moat, not a principle. Hiring one of the authors of SB 53's intellectual foundation is both an answer to that objection and a fresh invitation to it. A person who helped design the rules moving to a regulated company is the revolving door in its textbook form. Neither Anthropic's announcement nor Daniela Amodei's quote addresses that directly.

Cuéllar's side of the trade is clear enough. Five years running a think tank let him launch AI governance programs, but a think tank's influence ends at reports and persuasion. A frontier lab's policy chief can touch actual release conditions, usage policies, and government contract terms. When he says there's "no more consequential place to be shaping that work," that's the difference he's describing. The less flattering reading is that a career civil servant and jurist at the peak of his public standing moved to a company — and that the center of gravity in AI policy has moved with him.

There's also a party that plainly loses something here: the Long-Term Benefit Trust. The Trust is an independent body that holds Class T stock giving it the power to elect and remove Anthropic board members — it was designed to elect a majority of the board within four years — plus protective provisions requiring notice of major corporate changes. Trustees hold no Anthropic equity and share in no profits. That body just handed one of its members to management six months after recruiting him. The remaining trustees are former Federal Reserve chair Ben Bernanke, who joined in July, Center for a New American Security CEO Richard Fontaine, and chair Neil Buddy Shah. Founding trustees Kanika Bahl and Zachary Robinson concluded their terms back in January. When an oversight body turns over that fast, the continuity of the oversight itself becomes a fair question.

Nick Clegg, Joel Kaplan, Chris Lehane — how this job goes right and wrong

Big tech companies hiring an outside global-policy chief is a well-worn script with two very different endings.

The most famous case is Meta's Nick Clegg. The former UK deputy prime minister joined Facebook in 2018 and eventually became president of global affairs. He delivered real things: the Oversight Board that adjudicates content decisions, and functioning channels to European regulators. Then in January 2025, days before Trump's second inauguration, Meta replaced him with Joel Kaplan, a Republican who had been George W. Bush's White House deputy chief of staff for policy, as Chief Global Affairs Officer. The lesson is unsentimental: this job's occupant changes when the company's regulatory environment changes. When Meta needed the European language of legitimacy, it was Clegg. When it needed Washington Republican channels, it was Kaplan.

OpenAI's Chris Lehane is a different archetype. A Clinton-administration political communications operator who ran policy and communications at Airbnb from 2015 to 2021, fighting city regulators head-on, he joined OpenAI as Chief Global Affairs Officer in 2024 and now also serves as interim Chief Communications Officer. The logic there is to hire someone who has beaten regulators before. Anthropic's logic runs the other direction: it hired a rule-designer, not a brawler.

The best-case precedent is Microsoft. Brad Smith started as an in-house lawyer, became general counsel, and is now vice chair and president. A company scarred by 2000s antitrust litigation put legal and policy people at the very top and spent twenty years rebuilding its relationship with regulators. A meaningful share of the political stability Microsoft enjoys today in cloud and AI procurement traces back to that investment. The takeaway: policy organizations are measured in years, not quarters.

And there's a failure mode. Uber in 2017 loaded up on political stars including David Plouffe, but while the company's actual conduct stayed the same, famous political hires didn't neutralize its regulatory exposure. That's the only way this role reliably fails — when what the company does and what its policy chief says drift apart, the hire becomes a cost line. Applied to Anthropic, the test is specific: the company walked away from a roughly $200 million contract and went to court to keep a usage policy barring autonomous lethal weapons and mass domestic surveillance. As long as that policy holds, Cuéllar's words and the company's conduct match. If it bends, his credentials stop being an asset.

How rivals will push back

The most direct counter comes from OpenAI. It already has an expanded global policy team under Lehane, and through Leading the Future it is pushing federal preemption with $25 million of Brockman's own money in the pot. The frame available to OpenAI writes itself: keep arguing that pro-regulation positioning is anticompetitive, and treat this hire as evidence of regulatory capture. The fact that Anthropic recruited an appellate justice who helped shape California's frontier AI law is usable material for exactly that argument.

Google and Microsoft fight on a different layer, and their counterplay is closer to indifference than response. Both have run mature policy organizations for two decades with established faces in Kent Walker and Brad Smith, and both sell AI inside existing cloud procurement vehicles, which means government relations is already fused to the revenue pipe. Anthropic isn't there. It is instead trying to litigate its way out of exclusion from Defense Department procurement. That gap is the real backdrop to this hire.

Meta can poke at another nerve. Under Kaplan it has argued AI policy through open-source and free-expression framing, though that frame lost some force when Meta effectively stopped publishing weights with Muse Spark in April 2026. Even so, "a closed frontier lab is writing the rules" remains a live attack, and right now Anthropic is the easiest company in the industry to aim it at.

The counterparty people forget, though, isn't a competitor at all — it's the executive branch. Both the March designation and the June export controls were government-initiated. Reporting on the summer summary judgment hearing indicates Judge Lin questioned whether the record supported the government's designation at all, but no final ruling has issued, and there has been reporting that the administration asked the court to reimpose the designation. The government retains room to appeal or to apply pressure through other instruments. So the first real test of this job isn't rival messaging; it's whether the relationship can be rebuilt after the litigation ends.

The state-level front stays open too. SB 53 is in force in California, and other states are writing their own rules, including training-data disclosure requirements. The Senate once killed federal preemption of AI law 99–1, and the industry reopened that fight with super PAC money. Cuéllar helped build half of this terrain himself. For Anthropic that is a powerful card. For the other side, it's a precise target.

So what actually changes

If you buy AI for a company, read this as a procurement signal. The lesson of Anthropic's first half of 2026 is that frontier models can go dark for political reasons, not technical ones — a single export control order pulled two top-tier models away from every customer for roughly three weeks in June. That belongs in your vendor review: outage history driven by regulation or export control, and what your fallback path is. The fact that Anthropic elevated this function to the C-suite is itself a sign of a company that has been asked those questions the hard way.

For developers, nothing changes at the API surface. The indirect channel is usage policy stability. What Anthropic protected by forfeiting a roughly $200 million contract was its ban on autonomous weapons and mass surveillance uses — and that clause lives in the same family of documents as the acceptable-use terms every developer agrees to. Filling the top policy seat with a jurist means future revisions of those documents get drafted more precisely. Charitably, that's predictability. Less charitably, it means longer terms.

For investors, the thing to watch is the price of regulatory risk. Anthropic raised $30 billion in a Series G on February 12 at a $380 billion post-money valuation, and per CNBC its May round valued it close to $1 trillion, with a reported annualized revenue run rate of $47 billion at that point. A company at that scale spent the first half of the year with its largest single risk being neither a competitor nor a capability gap, but the federal government. Lobbying outpacing all of 2025 in six months, and $40 million routed to a political group, are that risk priced in dollars. This hire is one more executive-level line item on the same budget.

For ordinary users, there is no immediate change. What the story does reveal is a shape worth internalizing: over the next few years, what AI is permitted to do will be decided in courtrooms and legislatures more than on leaderboards. Whether Claude can be used in autonomous weapons, whether foreign nationals can access the top model tier, whether a state can impose transparency duties on frontier developers — every Anthropic controversy of 2026 has been one of those questions. And the person the company hired to answer them is a judge.

Put simply: Anthropic didn't create this role to improve its messaging. It created a C-suite seat to pull together a government it is suing, an export control regime that can reactivate at any time, a rulebook fragmenting across 50 states, and government relationships scattered across 12 cities. Filling it with someone who spent six months watching the company from its oversight trust is the single most unusual thing about the hire — and the single most contestable.

🥄 Three Things You're Probably Wondering

— So what does this mean for me? If you just use Claude, nothing changes today. Worth remembering, though, that a single export control order took two top-tier models offline for weeks in June. If you've wired a specific model deep into a workflow, it's cheap to verify your fallback now.

— Why is this happening now? In March the Defense Department labeled Anthropic a supply chain risk and shut it out of federal contracting; in June a Commerce export order stopped two models; in July and August the court heard cross-motions for summary judgment. Lobbying spend in six months already exceeded all of 2025. Policy risk is touching revenue and product directly, which is what a new C-suite seat is for.

— Is Anthropic ahead of its rivals here? On credentials it's an unusual card. OpenAI's Chris Lehane is a political and communications operator, Meta's Joel Kaplan came from a Republican White House, and Cuéllar is a former justice who helped design California's frontier AI law. Whether one hire closes the gap with Google and Microsoft, where government relations is already fused into how they sell, is too early to call.

Sources

Numbers are as of announcement and may change.