Dario Amodei

On March 3, 2003, while studying at the California Institute of Technology, Amodei published a signed article in The California Tech, ‘Anti-War Protest: Most Students Apathetic Towards World Politics.’ He opposed the impending Iraq War, criticized political apathy on campus, and urged students to take part in anti-war petitions and discussion. He argued that scientists also had civic responsibilities and should apply their analytical skills to public issues.⁠[1]

On March 5, he helped organize an anti-war event outside the campus Red Door Cafe as a member of Peaceful Justice at Caltech (PJC). The event included a petition, an open microphone, and information tables; groups supporting the war also attended to debate. The organizers aimed to encourage discussion and provide information. Amodei said they would continue peaceful protests and connect with other groups, while fellow organizer Ki Won Yoon proposed taking an anti-war petition to the Pasadena City Council to seek a resolution opposing the war.⁠[2]

On March 10, Amodei spoke at a Pasadena City Council meeting in California and read a petition opposing the invasion of Iraq. The petition expressed concern about scientific technology being used in warfare, advocated addressing the issues through the United Nations, and called for the protection of peaceful research that brought jobs and scientific progress to the community. The city’s archives retain a recording of that day’s special meeting.⁠[3][4]

Amodei studied as an undergraduate at Stanford UniversityStanford UniversityStanford University is a research university in California founded by Leland and Jane Stanford in 1885 and opened in 1891.Open the full entry and later pursued doctoral research in physics and biophysics at Princeton UniversityPrinceton UniversityPrinceton University is a private university and Ivy League member in Princeton, New Jersey. Chartered on October 22, 1746 as the College of New Jersey, it moved to Princeton in 1756 and adopted its present name in 1896. Undergraduate education leads to A.B. and B.S.E. degrees and emphasizes small-group discussion, independent research and residential college life. Graduate education spans the humanities, social sciences, natural sciences and engineering. As of October 2026, Christopher L. Eisgruber was president. Its 2026–27 operating budget was $3.407 billion, including $342 million for undergraduate aid.Open the full entry. He was selected as a Hertz Fellow in 2007. His research examined the collective behavior of neural circuits, combining statistical mechanics models with methods for recording neural activity.⁠[5][6]

After completing his doctorate, he conducted postdoctoral research at the Stanford University School of Medicine. In 2012, the Hertz Foundation awarded its Thesis Prize to him and two other recipients, recognizing his work on electrophysiological measurements and collective behavior in neural circuits.⁠[6][7]

Amodei recalled joining Baidu’s AI research team in late 2014, working on speech recognition with Andrew NgAndrew NgAndrew Ng is a British-born computer scientist and AI educator. A former director of the Stanford Artificial Intelligence Lab, he helped found Google Brain and Coursera, served as Baidu’s chief scientist from 2014 to 2017, and later founded DeepLearning.AI, LandingAI and AI Fund.Open the full entry and others. During this period, he began observing the relationship between improvements in performance and increases in model size, data volume, and training investment.⁠[8]

He was a co-author of the Deep Speech 2 paper released in December 2015. The research used end-to-end neural networks for English and Mandarin speech recognition and discussed how high-performance computing could improve training efficiency and support low-latency deployment.⁠[9]

Before joining OpenAI, Amodei was a senior research scientist at Google Brain. At OpenAI, he served as vice president of research, helped set the research direction, and led the development of large language models including GPT-2 and GPT-3.⁠[6][10]

In 2016, he co-authored Concrete Problems in AI Safety with Chris OlahChristopher OlahChristopher Olah (commonly known as Chris Olah) is a machine learning researcher and co-founder of Anthropic who works on neural network interpretability. He was previously a research scientist at Google Brain, led the Clarity team at OpenAI, and founded the scientific journal Distill.Open the full entry, Jacob Steinhardt, Paul Christiano, John Schulman, and Dan Mané. Addressing incorrect objectives, the cost of supervision, and undesirable behavior during learning, the paper identified research problems involving side effects, reward hacking, scalable oversight, safe exploration, and distributional shift.⁠[11]

In December 2020, OpenAI announced that he would leave after nearly five years to pursue a new research project with a small number of colleagues. The announcement recognized his work with Ilya SutskeverIlya SutskeverIlya Sutskever is a computer scientist who studied under Geoffrey Hinton, helped develop AlexNet and sequence-to-sequence learning, and is a co-founder and former chief scientist of OpenAI. As a board member in 2023 he took part in removing Sam Altman, and after leaving OpenAI in 2024 he founded Safe Superintelligence, which he has led as CEO since 2025.Open the full entry in jointly leading the organization’s research direction.⁠[10]

In 2021, Amodei co-founded AnthropicAnthropicAnthropic is an artificial intelligence research and product company founded in 2021 that develops the Claude models, assistant, and tools such as Claude Code, and operates as a public benefit corporation. Dario Amodei and Daniela Amodei are its co-founders, and its research covers model safety, interpretability, and steerability.Open the full entry with Daniela AmodeiDaniela AmodeiDaniela Amodei is an American executive who co-founded the artificial intelligence company Anthropic, where she serves as president and chairs the board. After early work in international development and politics, she held roles at Stripe and OpenAI before co-founding Anthropic in 2021 with her brother Dario Amodei and others.Open the full entry and others. He became chief executive officer, while Daniela became president; the two are siblings. The company initially focused its research on the reliability, interpretability, steerability, and use of human feedback in large AI systems.⁠[12][13]

In May of that year, the company announced a $124 million Series A financing round for research and prototype development. Amodei continued to oversee the company’s research direction and strategic vision.⁠[12][13]

On March 14, 2023, Anthropic expanded access to Claude. Available through a chat interface and a developer API, the assistant supported tasks including summarization, question answering, writing, search, and coding. The initial offering included Claude and Claude Instant.⁠[14]

In October 2024, Amodei published Machines of Loving Grace, discussing changes that powerful AI might bring to health, neuroscience, economic development, governance, and the meaning of work. He explicitly framed these ideas as hypotheses and predictions, acknowledging that experiments, data, and the pace of real-world processes would constrain progress.⁠[15]

The essay argued that taking risks seriously was compatible with anticipating positive outcomes from technology, and that realizing these benefits would require research, institutions, and social action.⁠[15]

In January 2026, he published The Adolescence of Technology, turning to the risks of powerful AI and society’s response. He argued for acknowledging uncertainty in predictions, avoiding quasi-religious approaches to technological risk, and making regulation as clear, prudent, and targeted as possible.⁠[16]

On February 26, Amodei issued a statement on Anthropic’s contract negotiations with the U.S. Department of War. He supported using AI for national defense but maintained exclusions for mass domestic surveillance and fully autonomous weapons without appropriate safeguards. He refused to remove those restrictions in order to accept a requirement allowing any lawful use.⁠[17]

In August, a federal court in San Francisco ruled that the government’s measures against the company were unlawful; another case remained pending. On September 3, Department of War officials continued to state publicly that the company’s supply-chain risk designation remained in effect.⁠[18][19]

On September 12, he published We Must Pace the Frontier, proposing three layers of action: embedded third-party evaluations, industry coordination, and international coordination. He argued that safety research and evaluations should keep pace with advances in model capabilities. Anthropic also committed to providing third-party evaluators with sustained, in-depth access.⁠[20][21]

The essay acknowledged that coordination among competitors raised antitrust issues and proposed narrow government exemptions for specific safety discussions. On September 18, four paying users sued Anthropic, OpenAI, SpaceXAI, and Google, alleging that coordination around slowing development harmed competition and the value of subscriptions. These were allegations in the plaintiffs’ complaint, not a finding that the companies had violated the law.⁠[20][22]

On October 27, 2026, The Builders, the interview series of biographer Walter IsaacsonWalter IsaacsonWalter Isaacson is an American biographer and journalist, born in New Orleans in 1952, and a history professor at Tulane University. He formerly led TIME magazine, CNN and the Aspen Institute. His biographies cover figures including Steve Jobs, Albert Einstein, Leonardo da Vinci, Jennifer Doudna and Elon Musk. He received the National Humanities Medal in 2023.Open the full entry, released his conversation with his sister Daniela Amodei; the program notes say it was taped on July 6, 2026. Dario said the range of skills needed to run a company is beyond any one person, that he is best at setting the big-picture strategy and research direction, and that Daniela runs the company day to day and handles recruiting. He also said Anthropic’s seven co-founders hold equal stakes; many people had warned them that seven co-founders would be a disaster, but all seven were still at the company.⁠[23]

Looking back on his time at OpenAI, he called GPT-2 the first real empirical test of the scaling laws and said he had been badly shaken by it: “My god, you just scale this thing up another factor of a million and it’ll be better than humans at everything.” He said about a tenth of the people around him reacted the same way, while most dismissed it because the model was not impressive at any single task. He also said that part of his inspiration for the original scaling laws came from the link in primate evolution between rapid growth in brain size and the emergence of language.⁠[23]

Asked whether he left OpenAI partly because of a breakdown of trust with Sam AltmanSam AltmanSam Altman is an entrepreneur and investor, and the co-founder and chief executive officer of OpenAI. He previously served as president of Y Combinator and founded the location-based social service Loopt. He also helped found Tools for Humanity and Merge Labs, has invested in energy and life sciences ventures, and has driven public discussion of AI governance, employment, and wealth distribution.Open the full entry, he said disagreements over safety at Anthropic get resolved because everyone believes they are held in good faith and that everyone is being honest; without that belief, disagreements cannot be resolved: “That’s why we left.” He also described speaking to the whole company for an hour every two weeks about the company, the field and the world, sessions employees call the “Dario Vision Quest” (DVQ), under two rules: nothing he says is an actual decision, and he is completely honest with the company.⁠[23]

On Claude’s training, he said the company had published a constitution for Claude a few months earlier; early on it had imagined governing the model with discrete rules, but found that a list of some 170 rules worked less well than teaching the model basic principles and letting it reason from them. On the dispute with the Pentagon, he said Anthropic had two concerns about autonomous weapons: that the technology is not yet reliable enough, and oversight and accountability, since one person could control a million drones that, unlike human soldiers, would not refuse unconstitutional orders. He said the company was not preventing the government from doing this but was unwilling to do it itself, that several proposals had since appeared in Congress to write those red lines into law, and that “the world was decisively on our side.” Asked about brain-computer interfaces such as Neuralink, he said AI should not be tied to any single vision of a good life and that he did not want to link his own brain to a computer.⁠[23]

Personal website: https://darioamodei.com/ (opens in a new window). X: Public account: https://x.com/DarioAmodei (opens in a new window). His personal website collects public essays, research links, and interviews, while Anthropic’s leadership page provides information about his current role.⁠[6][21][13]