Who Is Behind Claude: How Former OpenAI Staff Built the Anthropic Team

Anthropic grew out of an internal split at OpenAI and, within a few years, turned Claude into one of the most visible products on the AI market. Behind the model stands not one charismatic founder but a team with distinct roles: research, safety, operations and behavior tuning.

Where Anthropic came from

Anthropic was launched in 2021 by former OpenAI employees, including Dario and Daniela Amodei. They did not leave over personal drama but over a different view of how to build powerful artificial intelligence. The new company bet from the start on controllability, interpretability and model safety, not only on the race for flashy demos.

By April 2026 that strategy had paid off: Anthropic is now seen as OpenAI's main rival, and its valuation in the official funding round reached $380 billion. On the secondary market, according to the business press, investors pushed it even higher, to roughly $1 trillion.

Claude went viral not because of one polished release but because of its practical usefulness. Users came to like it for working with long texts and for analyzing correspondence, code and documents without noticeable loss of context. Claude took off especially loudly among developers: the separate Claude Code tool can read a codebase, edit files, run tests and bring tasks to a working state. The industry noticed too: Jensen Huang publicly called Claude a strong leap in coding and reasoning.

Who makes Claude

The article's main point is that Claude is backed not by an abstract "lab" but by a very concrete group of people with different skills. One sets the scientific trajectory, another keeps the company running, a third tries to understand the internal mechanics of the models, and a fourth is responsible for the assistant's character and behavioral style. Together it looks less like another chatbot and more like a product with its own philosophy.

• Dario Amodei — CEO and co-founder of Anthropic, former OpenAI vice president of research, who worked on GPT-2 and GPT-3.
• Daniela Amodei — president of Anthropic, responsible for operations, processes, partnerships and for making sure the company can withstand its own growth.
• Chris Olah — co-founder and interpretability researcher, who studies how a model's thinking is actually structured inside.
• Amanda Askell — a philosopher by training who works on personality alignment, that is, on making Claude helpful, honest, careful and polite.

This team has a composition that is rare in the AI market. Alongside the classic cult of the "strong engineer" are people who know how to build processes, set ethical boundaries and dissect a model as a complex organism. It is no accident that Anthropic is often described as a company whose product is built simultaneously on research, safety policy and very fine-grained behavior tuning. That is why Claude is seen not just as a powerful model but as a system with character.

Safety as strategy

Anthropic's public line is easy to read in Dario Amodei's texts and decisions. In October 2024 he published the essay Machines of Loving Grace, describing a scenario in which powerful AI helps medicine, education, the economy and governance rather than becoming a source of chaos. In January 2025 he wrote separately about DeepSeek, China and export controls, insisting that democratic countries must keep the lead in AI. And in April 2025 he published The Urgency of Interpretability: powerful models already exist, but we still understand poorly what exactly happens inside them.

"The elegance of machine learning is the elegance of biology."

That idea also captures the approach of Chris Olah, one of Anthropic's key researchers. For him a model is not simply an engineered system with understandable gears but a complex digital organism that has to be opened up and studied almost like a living structure.

The same approach showed up in the loudest conflict around Anthropic, in February 2026. The company discussed Claude's work with the Pentagon but refused to remove two restrictions from the contracts: a ban on mass domestic surveillance and a ban on fully autonomous weapons. This was not a PR pose but a demonstration that, for Anthropic, safety is part of the product, not an add-on to it.

What this means

Claude's story is not only the story of a successful model but of a well-assembled team. Anthropic shows that in the AI race the winner is not whoever simply ships another chatbot but whoever combines strong research, operational discipline and a clear stance on safety. For the market this is a signal: competition with OpenAI is no longer only about answer quality but about whom users and companies are ready to trust with real work tasks.