Dario Amodei is an Italian-American AI researcher and entrepreneur who co-founded Anthropic after leading safety and scaling-laws work at OpenAI.1 A Princeton-trained physicist, he helped demonstrate that larger models predictably improve with more compute and data, shaping today’s AI boom.2 As Anthropic’s CEO, Amodei now balances rapid capability advances with rigorous safety commitments, positioning the Claude model family as a transparent, controllable alternative in the race for powerful AI.3
Pioneering Scaling Laws at OpenAI
While directing research at OpenAI, Amodei co-authored landmark studies showing neural-network performance scales smoothly with model size, data and compute—insights that unlocked GPT-2 and GPT-3.2 He also led reinforcement-learning-from-human-feedback projects that proved large models can align better with human preferences.4 This body of work earned him a reputation as both a scientific driver and a safety champion inside the lab.
Building Anthropic for Safer AI
Departing OpenAI in 2021, Amodei founded Anthropic with his sister Daniela and fellow researchers to pursue "constitutional" AI systems that are steerable, interpretable and robust.3 Operating as a Public Benefit Corporation, Anthropic raised billions from Google, Amazon and others to train the Claude series on vast clusters while embedding safety checkpoints before, during and after training. The company’s voluntary government disclosures and red-team evaluations have since set industry benchmarks for responsible deployment.
Global Advocacy for AI Governance
Beyond research, Amodei briefs legislators, joins safety institutes and publishes essays—most notably “Machines of Loving Grace” — arguing that rigorous oversight is essential to unlock AI's upside without catastrophic risk.5 He advocates export-controlled chip supply chains, open-sourced evaluation protocols and an international “entente” among democracies to curb authoritarian misuse. TIME named him one of the 100 most influential figures in AI for elevating safety to a board-level priority.1