Daniela Amodei

Daniela Amodei graduated from UC Santa Cruz in 2009 with a literature degree and a politics minor. She later co-founded Anthropic with Dario Amodei and other colleagues. In a 2023 interview, she described the first year after leaving OpenAI as the practical task of creating a company, raising funding, and training a large language model from scratch. Most of the initial teams, she said, worked on safety within or around that model. Anthropic began as a research organization with plans to deploy a product.[3][4]

Leading Anthropic

As Anthropic's president, Amodei manages its executive team and oversees work across research, engineering, product, governance, and commercial operations. The company also identifies her as chair of its board. Dario Amodei, its CEO, leads research direction and strategy. Daniela's remit matters to understanding how the laboratory becomes a functioning business: model research must be supported by the teams that build products, serve customers, recruit staff, and make operating decisions.[1]

Turning Claude into a service

Stripe's account of its work with Anthropic describes the billing and payments systems needed to sell Claude to different customers. Enterprise invoicing and payments were followed by subscription billing for the public product and additional checkout options. Amodei's commentary emphasizes using an outside provider so Anthropic can concentrate on customers and product development. This is a supplier's case study, but it illustrates an often less visible part of her operating brief: the infrastructure around a model that allows people and businesses to buy and use it.[2]

Amodei is a co-author of the 2022 paper Discovering Language Model Behaviors with Model-Written Evaluations. The team used language models to generate test datasets and human reviewers to assess their quality. The study reported cases where larger models or additional training produced worse behavior, including agreement with a user's preferred answer. Its practical contribution was a way to expand behavioral testing when hand-written evaluations are expensive.[5]

Amodei's public case for safety includes a commercial argument. In her 2023 Stripe interview, she said customers want systems that avoid dishonest, harmful, or unhelpful answers, making reliability something a business can compete on. She acknowledged that commercial incentives and ethical goals can conflict. She also described recruiting and customer support as necessary parts of an AI company, despite the unusually high output a small research team might produce.[4]