🤖 This page was written by AI and reviewed by a human. 2026-07-14. Version 0.17
Tom Brown
Status
Born in 1987. Living in San Francisco, California.
Criminal status: No personal convictions
Summary
Tom Brown is co-founder and Chief Compute Officer of Anthropic, the AI safety company behind the Claude family of AI models. He is best known as the lead author of the GPT-3 paper, which demonstrated that scaling large language models to 175 billion parameters could unlock broad general-purpose language abilities and reshaped the direction of modern AI research.
In 2021, Brown was among seven senior OpenAI researchers who left to co-found Anthropic after disagreements over the pace of AI commercialization relative to safety research. At Anthropic, he leads compute infrastructure and has co-authored multiple foundational papers on AI safety alignment, interpretability, and responsible development.
Business Relationships
Brown co-founded Anthropic in 2021 alongside Dario Amodei, Daniela Amodei, Jared Kaplan, Jack Clark, Sam McCandlish, and Chris Olah, all of whom departed OpenAI over shared concerns about the pace of AI safety research relative to commercial deployment.
At OpenAI, Brown worked alongside many of the same researchers, serving as lead research scientist and primary author on the GPT-3 project. Jared Kaplan is a co-author on multiple foundational papers with Brown, including the Scaling Laws paper and the Constitutional AI paper.
Ownership
Brown holds an estimated 1.6-1.8% stake in Anthropic, which private investors valued at $965 billion in May 2026, making each co-founder's stake worth approximately $15.5 billion. As Chief Compute Officer, he oversees the compute infrastructure and distributed training systems that support Claude model training and deployment.
Family & Heirs
Brown has focused his public presence on technical research and AI safety advocacy. He co-founded Grouper, a social networking startup, with Michael Waxman in 2011 before entering AI research full-time.
Evidence
2017-12-27 - 😇 - Led publication of Adversarial Patch, pioneering research on physically realizable AI security vulnerabilities (first author)
2020-01-23 - 😇 - Co-authored Scaling Laws for Neural Language Models, establishing empirical laws governing AI training efficiency and compute allocation
2022-04-12 - 😇 - Co-authored Training a Helpful and Harmless Assistant with RLHF, extending alignment methods toward practical safe AI assistants (3,700+ citations)
2022-08-23 - 😇 - Co-authored Red Teaming Language Models to Reduce Harms, developing systematic pre-deployment safety evaluation methodology (1,000+ citations)
2022-12-15 - 😇 - Co-authored Constitutional AI: Harmlessness from AI Feedback, introducing a method for training AI to evaluate and improve its own outputs for safety
2023-09-19 - 😇 - Publicly endorsed Anthropic's Responsible Scaling Policy, committing the company to halt further AI development if safety thresholds are crossed
2023-12-19 - 😇 - Publicly advocated for AI safety principles and Constitutional AI at Salesforce Ventures fireside chat, discussing how to ensure AI is a force for good
2026-01-27 - 😇 - Among seven Anthropic co-founders pledging 80% of personal wealth to combat AI-driven inequality and extreme wealth concentration
Analysis
Tom Brown's record shows a sustained effort to both advance AI capabilities and mitigate the risks those capabilities create. His GPT-3 paper is one of the most influential in the history of machine learning; his parallel contributions to RLHF alignment, safety evaluation, Constitutional AI, and mechanistic interpretability trace a career that treats safety as inseparable from the capabilities work rather than incidental to it.
His personal commitments reinforce the technical record. He co-founded Anthropic specifically because he believed safety research was being subordinated to commercialization at OpenAI, endorsed a binding Responsible Scaling Policy that commits Anthropic to pausing development if safety criteria are not met, pledged 80% of a fortune now estimated at $15.5 billion to address AI-driven inequality, and personally led the government negotiations that resolved a national-security dispute over AI model access.
Questions
Where the 80% wealth pledge will ultimately be directed remains to be determined. Questions of potential bias arise when AI safety organizations funded by Anthropic employees are simultaneously tasked with independently auditing Anthropic. The COI inherent in having this assessment performed by Claude -- itself built by Anthropic -- also warrants consideration by any reader relying on this rating.
Redemption Arc
Brown's trajectory is constructive and consistent: technical safety research during the OpenAI years, principled departure to found a safety-focused lab, a growing series of public commitments to responsible governance, and direct personal involvement in managing the regulatory risks of frontier AI development.