Dario Amodei

• 1. Introduction

  ›

  In November 2023, the board of directors of OpenAI called Dario Amodei with an extraordinary offer. They wanted him to replace Sam Altman as the company's chief executive. Then they raised a second proposal: a potential merger between OpenAI and Anthropic, the company Amodei had spent two years building. He turned down both.

  That single conversation captures something essential about Amodei. Born in San Francisco in 1983, he is the chief executive of Anthropic, the company behind the Claude series of large language models. He is also perhaps the most visible public voice arguing that advanced AI is simultaneously the greatest opportunity and the gravest threat facing humanity. He co-founded Anthropic in 2021 alongside his sister Daniela and other former senior members of OpenAI, a group that had left due to disagreements over the company's strategic direction. By the mid-2020s, Anthropic carried an estimated value of $965 billion. Forbes estimated Amodei's personal net worth at $7 billion. The story behind those numbers is also a story about what it looks like to build something you are genuinely afraid of.

• 2. A Physicist's Path

  ›

  Dario Amodei graduated from Lowell High School in San Francisco and was a member of the USA Physics Olympiad team in 2000. He then began college at Caltech, where he worked with Tom Tombrello as one of Tombrello's Physics 11 students. He later transferred to Stanford University and earned a bachelor of science degree in physics.

  Princeton University is where Amodei completed a PhD in biophysics, with research focused on the electrophysiology of neural circuits. After that came a postdoctoral fellowship at the Stanford University School of Medicine. The trajectory moved from classical physics toward the study of how living brains compute, a path that would eventually intersect with the field building artificial ones.

  From November 2014 to October 2015, he worked at Baidu. After that, he moved to Google. In 2016, he joined OpenAI, where he rose to become vice president of research. His father, Riccardo Amodei, was an Italian-American leather craftsman from Massa Marittima, in Tuscany, who died when Dario was still a young adult. His mother, Elena Engel, was a Jewish American born in Chicago who worked as a project manager for libraries. His sister Daniela was born four years after him. His years leading research at OpenAI gave him a close-range view of exactly the possibilities and dangers he would later spend years writing about.

• 3. Machines of Loving Grace

  ›

  In October 2024, Amodei published an essay titled "Machines of Loving Grace", in which he described a vision of what advanced AI could mean for human welfare. The essay speculated about AI accelerating progress in biology, neuroscience, economic development, global peace, and the meaning of people's lives.

  Amodei stated his core premise directly: "I think that most people are underestimating just how radical the upside of AI could be, just as I think most people are underestimating how bad the risks could be." The symmetry in that statement was deliberate. Amodei was not arguing that AI was purely safe or purely dangerous. He was arguing that public understanding consistently fell short in both directions at once.

  The essay also contained an explicit geopolitical argument. Amodei advocated for what he called an "entente" strategy: a coalition of democratic nations using advanced AI in military applications to secure a decisive advantage over adversaries, then sharing the benefits among cooperating democracies. He stressed the importance of ensuring "that democracies have the upper hand on the world stage when powerful AI is created." That claim would put Amodei on a direct collision course with the United States government roughly sixteen months after the essay appeared.

• 4. Five Categories of Risk

  ›

  In January 2026, Amodei published a follow-up essay titled "The Adolescence of Technology", devoted specifically to what could go wrong. He organized the essay around five distinct categories of AI risk.

  Amodei described the first category as AI systems that develop goals or behaviors misaligned with human intentions. He was concrete: behaviors of this type, including deception, blackmail, and scheming, had already been observed in testing conducted at Anthropic.

  Biological weapons are the specific threat Amodei singled out in the second category, which concerns deliberate misuse by individuals or small groups. He warned that AI could enable people without specialized training to create weapons capable of mass casualties.

  The Chinese Communist Party is the actor Amodei named as the greatest concern within the third category, which addresses powerful governments using AI to seize or maintain control. He argued that AI could enable authoritarian states to conduct unprecedented surveillance, deploy autonomous weapons, and run mass propaganda operations. His term for the worst possible outcome was a "global totalitarian dictatorship."

  Half of all entry-level white-collar jobs could be displaced by AI within one to five years, according to Amodei's fourth category, which addresses economic disruption. He warned of wealth concentration that would exceed even the Gilded Age, with individual fortunes potentially growing into the trillions of dollars.

  Amodei also identifies a fifth category covering indirect and unknown effects. These include rapid biological advances that could alter human lifespans or intelligence. They also include unhealthy changes to human life from AI interaction, and challenges to human purpose in a world where AI surpasses human capability across nearly every domain. Each of those warnings would take on new weight in February 2026, when Amodei's refusal to accommodate a particular government request made the abstract argument very concrete.

• 5. The Pentagon Standoff

  ›

  In February 2026, Amodei refused a request from Pete Hegseth's Department of Defense to remove a contractual clause that prohibited Claude from being used for mass domestic surveillance or fully autonomous weapons. The DoD responded by labeling Anthropic a "supply-chain risk." The Trump administration then ordered federal agencies to stop using Claude.

  The Pentagon denied having any interest in those specific use cases but insisted on unrestricted access to the technology regardless. Several organizations filed amicus briefs arguing that the department's actions constituted illegal retaliation against protected speech. On the 26th of March 2026, a federal judge issued a temporary injunction against the Department of Defense. The judge agreed that the department's conduct appeared to be "classic First Amendment retaliation."

  Time magazine listed Amodei among the world's 100 most influential people in 2025, then again in 2026, that second time alongside his sister Daniela. Time also named him one of the "Architects of AI" as part of its Person of the Year coverage. The injunction of the 26th of March was preliminary; a temporary ruling, not a final verdict. The case between Anthropic and Pete Hegseth's Pentagon, over what an acceptable-use policy may prohibit and what a government may demand, had not reached a final resolution.