Dario Amodei

In May 2026 the US government gave Anthropic a short deadline to drop two restrictions from a Department of Defense contract: no systems for domestic mass surveillance of Americans, and no fully autonomous weapons. Anthropic refused, the Trump administration answered with a supply chain risk designation, a mechanism previously used only against foreign entities such as Kaspersky Labs, and Amodei called holding the line an exercise of a constitutional right. "Disagreeing with the government is the most American thing in the world."¹

Amodei is the co-founder and CEO of Anthropic, an artificial intelligence company that builds large language models, sold commercially under the name Claude and licensed to businesses, developers, and government agencies. Anthropic is one of a small number of firms, alongside OpenAI and Google DeepMind, developing frontier AI systems at the largest scale. Amodei trained as a physicist and biophysicist before moving into machine learning research, and founded Anthropic around the idea that AI safety research must be a co-equal priority with capability research, not an afterthought.

Background

Amodei was born in 1983 and grew up in the Mission District of San Francisco. His father, Riccardo Amodei, was an Italian leather craftsman; his mother, Elena Engel, worked as a project manager for libraries. He attended Lowell High School and competed in the USA Physics Olympiad in 2000, then studied physics at Caltech before transferring to Stanford, where he finished his bachelor's degree. He earned a PhD in biophysics from Princeton in 2011 as a Hertz Fellow, advised by Michael Berry and William Bialek, with a dissertation on the collective behavior of neural circuits. His father died in 2006 from a rare illness that became largely treatable within a few years of his death. Amodei has cited that experience as the source of a dual urgency that has shaped his career: a wish to accelerate technologies capable of curing disease, paired with an insistence on getting the safety of those technologies right.²

Getting started

After a postdoctoral position at Stanford applying machine learning to biomedical data, Amodei joined Baidu in November 2014, co-leading work on the Deep Speech 2 speech-recognition system. There he and colleagues observed that simply making neural networks larger and training them on more data produced smooth, predictable performance gains, a pattern later known as AI scaling laws.² He briefly worked at Google Brain, co-authoring "Concrete Problems in AI Safety" (2016), an influential early paper in the field. He joined OpenAI in 2016, leading its safety team, becoming research director in 2018, and later vice president of research, a role in which he led development of GPT-2 (2019) and GPT-3 (2020). With Paul Christiano at OpenAI, he also co-developed reinforcement learning from human feedback (RLHF), the technique later used to align ChatGPT and Claude. Amodei left OpenAI in December 2020 over disagreements about its direction and pace of safety work, saying it is "incredibly unproductive to try and argue with someone else's vision." In January 2021 he co-founded Anthropic with his sister Daniela Amodei, who became its president, along with roughly a half-dozen other OpenAI alumni.²

What he built

Anthropic raised about $124 million in its first funding round in May 2021 and grew from a small founding team into a company employing thousands. It released its first Claude models in 2023 and has since shipped successive generations, competing directly with OpenAI's GPT models and Google's Gemini models. Its safety research centers on Constitutional AI, a technique that trains models against a written set of principles, extending the RLHF approach Amodei helped originate. Amazon and Google both became large investors and cloud partners, with Amazon's total investment reaching roughly $8 billion by late 2024 and Google reportedly holding a stake in the mid-teens percentage range. Anthropic's valuation rose quickly through successive rounds, from about $61.5 billion in March 2025 to $183 billion that September to a reported $965 billion in a Series H round in May 2026, with annualized revenue reported to have crossed the tens of billions of dollars by mid-2026.² Anthropic also became a supplier of custom models to US intelligence and defense agencies, describing itself as the first AI company to deploy models on the government's classified cloud, with deployments Amodei has said span cyber operations and combat support.¹

How he operates

Amodei describes Anthropic's approach as leading at the frontier of AI capability while treating safety as inseparable from that work, not a constraint applied afterward. That produced concrete limits he says Anthropic will not cross even with large government contracts at stake: no systems for domestic mass surveillance of Americans, and no fully autonomous weapons that fire without a human in the loop on targeting or engagement. On surveillance, he argues the technology has outrun the law, that Fourth Amendment jurisprudence has not caught up with what AI-scale data analysis makes possible, and that Congress, not a private company, should ultimately settle the question.¹ On autonomous weapons, he distinguishes a values objection from a reliability objection, arguing today's AI systems are not yet predictable enough to be trusted with lethal decisions independent of human oversight.¹

Where things stand

As of mid-2026, Amodei remains CEO of Anthropic, now one of the most valuable private technology companies in the world and one of the most deeply integrated AI suppliers to the US government. Amodei has said the supply chain risk designation, as described by administration officials, overstated its legal scope, applying only to Anthropic's role within military contracts, and that Anthropic would challenge any formal action in court.¹ He has continued publishing on AI policy, including a 2026 essay arguing frontier AI now requires binding, FAA-style pre-release testing, and warning AI-driven job displacement may be severe enough to require new income supports.

Key facts

• Born 1983 in San Francisco's Mission District; PhD in biophysics from Princeton (2011), advised by Michael Berry and William Bialek.
• Co-discovered AI scaling laws empirically at Baidu in 2014-2015; later led GPT-2 and GPT-3 development at OpenAI.
• Co-developed reinforcement learning from human feedback (RLHF) with Paul Christiano, the technique underlying ChatGPT and Claude.
• Co-founded Anthropic in January 2021 with sister Daniela Amodei after leaving his VP of Research role at OpenAI in December 2020.
• Anthropic's valuation reportedly rose from roughly $61.5 billion in March 2025 to $965 billion in a May 2026 Series H round, backed by Amazon and Google.
• Maintains two public red lines, domestic mass surveillance and fully autonomous weapons, which contributed to a May 2026 US supply chain risk designation against the company.