Anthropic hires Tino Cuéllar: the governance layer of the AI race

Anthropic announced on August 4 that Mariano-Florentino (Tino) Cuéllar will become the company's first Chief Global Affairs Officer. He will lead work on public policy, international engagement and government relationships. Cuéllar previously led the Carnegie Endowment for International Peace, served on the California Supreme Court and worked at the intersection of technology, security and public policy at Stanford.

This may look like a conventional executive appointment. The larger signal is that AI companies are no longer competing only on models, GPUs and benchmarks. They are also building positions on how advanced AI should work with governments, regulators and society.

Why now?

Anthropic says Cuéllar will work with governments on the economic, security and social questions raised by rapid AI change. The company frames democratic influence over AI's direction and broad access to its benefits as priorities.

The role is broader than tracking individual laws. As AI enters public services, defense, education and critical business operations, questions around the model become just as important as capability: Who decides? Which risks are acceptable? Who is accountable when a system fails? How should rules travel across borders?

Institutions become part of the product

An AI company's public-policy team can look like a communications function from the outside. In practice, model access, safety reporting, data use, government contracts and crisis response form one governance chain.

Cuéllar's background across law, technology, international security and public institutions matters for that reason. At Anthropic's scale, technical decisions eventually meet decisions between institutions and countries. “What can the model do?” is joined by “Under what conditions should it be used?”

Companies should not replace the state

There is an important boundary here. Building a strong public-policy function does not mean AI companies should make public decisions themselves. Companies can provide expertise, technical evidence and risk analysis; legitimacy, oversight and final public decisions should remain with democratic institutions.

That distinction matters because AI development can move faster than regulation and public capacity. Companies need internal safety frameworks, but those frameworks cannot replace independent oversight or public debate.

The lesson for solo builders

This is not only a story about large companies. A small product team should ask similar questions early: What data goes to the model? What do users need to know? Who checks an incorrect output? Which steps can an agent take alone, and which require human approval?

The answers will not fit into one policy document. But data boundaries, permissions, irreversible actions and failure modes should be designed into the product. Choosing the rules around a model is part of the work, just like choosing the model itself.