Tom Brown (Tom Brown), an Anthropic co-founder, is reported to have taken the lead during a period of tensions with the White House and secured looser export limits on the company’s latest artificial intelligence models.

On July 1, Business Insider reported that Brown negotiated with the U.S. Commerce Department and played a role in lifting export restrictions on Anthropic’s flagship models, Phaible5 and Mythos5.

The episode shows Brown’s standing inside Anthropic has broadened beyond research and infrastructure into government relations. With chief executive Dario Amodei (Dario Amodei) not personally easing tensions, the comparatively low-profile co-founder stepped in as a negotiating channel. Anthropic’s head of public policy, Sarah Heck (Sarah Heck), also joined the talks, but Wired reported that a letter from Commerce Secretary Howard Lutnick (Howard Lutnick) on export controls was delivered directly to Brown.

Brown leads compute at Anthropic, securing the chips and infrastructure needed to train and run the company’s AI models. Anthropic uses Amazon’s Trainium, Nvidia graphics processing units and Google’s TPUs, distributing workloads across different chips depending on the task. This spring, as soaring demand for Claude Code increased pressure from a compute shortage, the company also announced a multibillion-dollar contract to run models at SpaceX’s large data centre Colossus.

Brown’s career can be summed up as moving between Silicon Valley startups and AI research. After graduating from the Massachusetts Institute of Technology, he worked at a language education startup and a mobile advertising company. In Y Combinator’s 2012 batch, he co-founded the dating startup Grouper. Michael Waxman (Michael Waxman), who worked with him at the time, described him as a good person, kind, sincere and a direct leader. He also said Brown had an ability to build bridges between different job functions.

After leaving Grouper, Brown shifted toward AI. He said on a Y Combinator podcast last year that he lacked the skills to jump straight into AI research at the time, and taught himself for several months using online courses, Kaggle projects and linear algebra textbooks. He also said he bought GPU chips with Y Combinator credits and used them for studying.

Brown later joined OpenAI in its early stage. He has recounted that soon after OpenAI’s launch, he contacted Greg Brockman (Greg Brockman) saying he wanted to help in any way and would do it even if it meant cleaning floors. He spent the first nine months on engineering work outside machine learning, but later became a key figure in AI research.

Brown contributed to a 2017 paper that laid the groundwork for reinforcement learning from human feedback. In 2020, he was listed as a leading author on research that established scaling laws for large language models and on the paper that introduced GPT-3. The paper showed that large language models pre-trained on vast data could perform new tasks in a human-like way with only a few examples, without fine-tuning. The GPT-3 paper in particular said he led the engineering work. His Google Scholar citation count now exceeds 140,000.

Brown has also publicly shared his view on the pace of AI development. In a 2024 co-founder discussion, he said he initially did not think AI would change society so quickly but changed his mind over time. "I also thought at first it wouldn’t be that fast," he said. "It changed over time. So I empathise with that perspective," he added.

For Anthropic, the outcome of the talks has become more important. The company is reviewing an initial public offering, and a listing, if completed, could help provide employee liquidity and fund expansion. In this context, Brown being deployed for sensitive negotiations in place of Amodei is also read as a sign of a deeper management bench.

The expansion of Brown’s role also ties into competition among AI companies. In the race for cutting-edge models, research results alone are not enough. Chip procurement, infrastructure contracts and regulatory responses must run in parallel, and Brown stands at that intersection. The structure has him leading technology development while negotiating with the government and coordinating deals with multiple semiconductor and cloud providers. It points to the possibility that his role will keep growing as Anthropic pursues a listing and expansion.