Dario Amodei

Training and early research

Dario Amodei earned a doctorate in biophysics from Princeton University as a Hertz Fellow and was a postdoctoral scholar at the Stanford University School of Medicine^[1]. He later worked at Google Brain as a senior research scientist^[1].

He is the first-listed author of Deep Speech 2, a December 2015 paper on an end-to-end deep learning approach to speech recognition in English and Mandarin^[15]. He is also the first author of Concrete Problems in AI Safety, a June 2016 paper co-written with Chris Olah, Jacob Steinhardt and others^[4].

Research leadership at OpenAI

Amodei served as vice president of research at OpenAI, where he led the development of large language models such as GPT-2 and GPT-3^[1]. His own site describes him as a co-inventor of reinforcement learning from human feedback^[1]. He is a co-author of Deep Reinforcement Learning from Human Preferences, a 2017 paper by Paul Christiano, Jan Leike, Tom Brown and others^[5].

He is also a co-author of Scaling Laws for Neural Language Models, published in January 2020 with Jared Kaplan, Sam McCandlish and others^[6]. In May 2020 he was among the authors of Language Models are Few-Shot Learners, the paper that introduced GPT-3^[7].

Founding Anthropic

Amodei left OpenAI and co-founded Anthropic with his sister Daniela Amodei, and TechCrunch reported the company's launch and $124 million in funding in May 2021^[2]. At that time Anthropic said it was primarily focused on research towards building large AI systems that are steerable, interpretable and robust^[2]. Amodei is the company's chief executive^[1][2].

Anthropic is a public benefit corporation whose stated mission is the responsible development and maintenance of advanced AI for the long-term benefit of humanity^[3]. Its board is elected by stockholders together with a Long-Term Benefit Trust^[3].

In September 2026 Anthropic said staff from Accenture would work inside the company to scrutinise its models and staff, making Accenture its first embedded evaluator under Amodei's plan to place third-party safety evaluators inside AI labs^[14]. TechCrunch reported that both companies expected to invest at least $1 billion in the project over five years^[14].

Essays and policy positions

In his October 2024 essay Machines of Loving Grace, Amodei described what a world with powerful AI could look like if things go right^[8]. He described such AI as a "country of geniuses in a datacenter" and guessed that it could compress the next 50 to 100 years of biological progress into 5 to 10 years^[8].

In January 2025 he published On DeepSeek and Export Controls, in which he argued that DeepSeek's model releases made US export controls on chips to China more important, not less^[9]. In The Urgency of Interpretability, published in April 2025, he wrote that Anthropic had set a goal of interpretability that can reliably detect most model problems by 2027^[10]. His site also lists a 2025 New York Times op-ed titled AI Needs Basic Transparency Rules^[1].

His January 2026 essay The Adolescence of Technology sets out five categories of risk from powerful AI: autonomy risks, misuse for destruction, misuse for seizing power, economic disruption and indirect effects^[11]. He wrote that the risks can be overcome if people act decisively and carefully^[11]. In Policy on the AI Exponential, published in June 2026, he discussed five policy areas, including regulation and public safety, macroeconomics and tax policy, and geopolitics^[12].

In We Must Pace the Frontier, published in September 2026, he proposed that frontier AI companies give embedded third-party evaluators ongoing access to verify safety practices, and wrote that Anthropic was committing to this step unilaterally^[13]. The essay also calls for democratic coordination on safety standards and, more cautiously, for global pacing that would require cooperation with China^[13].

Known for

• Co-founding and leading Anthropic
• Research leadership on GPT-2 and GPT-3 at OpenAI
• Co-authoring work on RLHF and scaling laws
• Essays including Machines of Loving Grace and The Adolescence of Technology

Sources

1. Dario Amodei's website: biography and list of writing — checked 6 Nov 2026
2. TechCrunch: Anthropic is the new AI research outfit from OpenAI's Dario Amodei, May 2021 — checked 6 Nov 2026
3. Anthropic: Company page — checked 6 Nov 2026
4. arXiv: Concrete Problems in AI Safety, June 2016 — checked 6 Nov 2026
5. arXiv: Deep reinforcement learning from human preferences, June 2017 — checked 6 Nov 2026
6. arXiv: Scaling Laws for Neural Language Models, January 2020 — checked 6 Nov 2026
7. arXiv: Language Models are Few-Shot Learners, May 2020 — checked 6 Nov 2026
8. Dario Amodei: Machines of Loving Grace, October 2024 — checked 6 Nov 2026
9. Dario Amodei: On DeepSeek and Export Controls, January 2025 — checked 6 Nov 2026
10. Dario Amodei: The Urgency of Interpretability, April 2025 — checked 6 Nov 2026
11. Dario Amodei: The Adolescence of Technology, January 2026 — checked 6 Nov 2026
12. Dario Amodei: Policy on the AI Exponential, June 2026 — checked 6 Nov 2026
13. Dario Amodei: We Must Pace the Frontier, September 2026 — checked 6 Nov 2026
14. TechCrunch: Anthropic's first embedded evaluator is Accenture, September 2026 — checked 6 Nov 2026
15. arXiv: Deep Speech 2: End-to-End Speech Recognition in English and Mandarin, December 2015 — checked 6 Nov 2026