Jared Kaplan

Jared Kaplan is Anthropic's chief science officer and, since October 2024, its Responsible Scaling Officer, the executive who decides whether a model has met the safeguards the company's Responsible Scaling Policy requires before it is trained further or released. He took the role over from co-founder Sam McCandlish.

He spent roughly fifteen years as a theoretical physicist before touching machine learning. He studied physics and mathematics at Stanford, wrote a Harvard PhD on holography under Nima Arkani-Hamed (2009), held a postdoc at SLAC and Stanford, and joined the Johns Hopkins faculty in 2012, working on quantum gravity and conformal field theory. He remains on the Johns Hopkins faculty, where he has taught the foundations of deep learning.

At OpenAI, while still a professor, he led "Scaling Laws for Neural Language Models" (January 2020), which found that a language model's loss falls as a smooth power law in parameters, data and compute. The paper gave labs a way to budget for bigger models, and Kaplan was among the authors of the GPT-3 paper four months later and of OpenAI's 2021 Codex paper. He left with Dario Amodei's group at the end of 2020. In Anthropic's first months, he told Wired, all the founders did technical work, and he also ended up running payroll "because, well, someone had to do it."

At Anthropic he is the senior author of "Constitutional AI" (December 2022), the method of training a model to critique and revise its own answers against written principles, and he heads the company's technical research. In a December 2025 interview with The Guardian he said humanity would face a decision between 2027 and 2030 about whether to let AI systems train their successors.