Anthropic’s founding story is one of the most consequential in the history of artificial intelligence. Seven researchers who helped build the most powerful AI systems in the world walked away because they were worried about what those systems might become. This is the complete history.
The OpenAI Origins
By 2020, OpenAI had produced GPT-3 — a 175-billion-parameter language model demonstrating qualitatively new capabilities. Dario Amodei, VP of Research, and several colleagues were growing increasingly concerned: what happens when these systems become significantly more capable? The company’s “capped-profit” structure and commercial partnerships with Microsoft were creating tensions with pure safety research.
The Precita Park Meetings
In spring 2021, senior OpenAI researchers began meeting in Precita Park, a neighborhood park in San Francisco’s Bernal Heights. These conversations crystallized around a founding team: Dario Amodei (CEO), Daniela Amodei (President), Jared Kaplan (CSO), Chris Olah, Tom Brown, Sam McCandlish (CTO), and Jack Clark. All seven had been at OpenAI. All seven left within a compressed time period in mid-2021.
The Founding
Anthropic was incorporated in 2021 as a Public Benefit Corporation (PBC) — a legal structure that formally embeds a social mission alongside profit objectives. The name “Anthropic” (relating to human existence) reflects the mission: building AI safe and beneficial for humanity. Early funding: $124 million seed from Spark Capital.
Constitutional AI
Anthropic’s most significant research contribution: Constitutional AI — training models to follow written principles rather than relying solely on human feedback at every step. The “constitution” is a list of principles Claude upholds: honesty, avoiding harm, respecting user autonomy. This creates more consistent safety behavior across a wider range of situations.
Growth and Current Status
Major investments from Google ($2B) and Amazon (up to $4B) validated Anthropic’s trajectory. By 2026, Anthropic is valued at over $61 billion. Claude competes directly with GPT-4o and Gemini as one of the three most capable AI assistants in the world. An IPO targeting late 2026 at $400-500B is widely expected.
Frequently Asked Questions
Who founded Anthropic?
Seven former OpenAI researchers: Dario Amodei (CEO), Daniela Amodei (President), Jared Kaplan (CSO), Chris Olah, Tom Brown, Sam McCandlish (CTO), and Jack Clark.
Why did the Anthropic founders leave OpenAI?
Growing concerns about AI safety practices and tensions between commercial pressures and rigorous safety research.
Leave a Reply