Women Building the AI Era, Part 5- Daniela Amodei

In 2021, eight people left OpenAI within weeks of each other. Among them were siblings Dario and Daniela Amodei, both vice presidents at the company, who had grown convinced that OpenAI was racing to commercialize its technology faster than its safety work could keep up.

They didn't just start a rival lab. They built a company, Anthropic, whose entire legal structure was designed to make caution and its business model the same thing. Dario became CEO, running research. Daniela became president, running the revenue, sales, and operations that actually pay for that research.

Five years later, Anthropic's Claude is used inside banks, hospitals, and government agencies that chose it specifically because of that safety reputation, and the company's revenue reached a reported 30 billion dollar annual run rate faster than almost any software company in history. Daniela Amodei didn't just help found an AI safety company. She helped prove that safety could be sold.

A Split Built Around One Disagreement

Before Anthropic, Daniela Amodei was OpenAI's vice president of safety and policy. Dario was vice president of research. Both, along with six colleagues, left in 2021 over what several later described as a widening gap between how fast OpenAI wanted to ship products and how carefully it was testing them first.

They weren't the only ones who left. Six other OpenAI researchers and executives departed alongside the Amodei siblings, including Jared Kaplan, who became Anthropic's chief science officer, and Tom Brown, a lead author of the research paper behind GPT-3 who now leads the technical organization that manages Anthropic's compute. It wasn't a resignation. It was closer to a team walking out together to run the same experiment under stricter rules.

Rather than simply criticizing OpenAI's approach from the outside, they built Anthropic to make the opposite bet structurally unavoidable. The company incorporated as a Delaware public benefit corporation, a legal structure that lets its board weigh the company's stated mission against shareholder returns, rather than shareholder returns alone. On top of that, Anthropic created the Long-Term Benefit Trust, an independent panel with the power to elect and eventually control a majority of the company's board, made up of people with backgrounds in AI safety, national security, and public policy rather than company insiders.

Think of it like a company that hires its own referee before the game starts, then hands that referee a growing say over who gets to keep coaching the team. Most companies let their board answer only to shareholders. Anthropic built a second, non-financial constituency directly into its own governance, specifically so that a bad quarter couldn't be used to argue the safety mission should bend.

Did You Know?

The Long-Term Benefit Trust doesn't hold any economic stake in Anthropic. Its members can't profit if the company's valuation rises, which was deliberately designed so that whoever controls part of Anthropic's board has no financial incentive to loosen the company's safety commitments under pressure.

Turning a Values Statement Into a Sales Pitch

Daniela Amodei's job is to run everything that keeps that structure functioning as an actual business: finance, enterprise sales, partnerships, recruiting, and policy. Under her, "we're the careful one" became a genuine market position rather than just a mission statement, particularly with customers who had the most to lose from an AI system behaving unpredictably.

That positioning helped Anthropic land customers in regulated, risk-averse industries, and by 2026 the company counted more than a thousand enterprise clients each paying over a million dollars a year, more than double the figure from just months earlier. Government agencies, banks, and healthcare providers don't typically move fast on new technology; Anthropic's safety-first branding gave them a reason to move faster than they otherwise would have, since the same kind of enterprise trust that makes an AI assistant usable inside a codebase is the trust that makes it usable inside a hospital record system.

The business results were dramatic. Anthropic's revenue run rate reportedly climbed from around 1 billion dollars at the end of 2024 to roughly 30 billion dollars by April 2026, a pace of growth that took established software giants decades to reach.

Interesting Fact: Anthropic reportedly reached its 30 billion dollar revenue run rate with a notably lean headcount for a company at that scale, keeping employee numbers in the low thousands rather than the tens of thousands typical of software companies with comparable revenue.

When the Strategy Meets the Competition

A safety-first strategy is easy to hold onto when nobody is racing you. It's harder to hold onto when rivals with fewer scruples are winning defense contracts you'd also like to win.

In February 2026, Anthropic rewrote a core piece of its own safety framework, the Responsible Scaling Policy, removing a previous commitment not to release a model unless the company could first guarantee adequate risk mitigations were in place. The company's public explanation was that some of its strongest safety commitments only made sense if competitors matched them, and that falling behind entirely would itself become a safety risk, not just a business one. Reporting at the time linked the shift partly to pressure to compete for the kind of government and defense contracts that rivals including OpenAI and xAI had already secured.

That pressure is only going to grow as Claude moves from answering questions to taking actions on a company's behalf, since an agent that can act independently inside a government system is a much bigger safety and sales question than a chatbot that only talks. Whether loosening the Responsible Scaling Policy is a principled recalibration or a strategy quietly giving ground to competitive pressure is a genuinely open question, and it's the one Daniela Amodei's job increasingly depends on answering correctly. Anthropic's entire commercial identity rests on being the company that won't cut corners. The moment customers start to doubt that, they have alternatives with fewer restrictions and, potentially, faster new features.

Knowlegic Perspective

It would be easy to read Anthropic's story as proof that doing the right thing and building a hugely valuable company are simply compatible, and Daniela Amodei's operating side of the business is genuine evidence for that case. Enterprise and government customers really did choose Anthropic partly because of its safety reputation, and that reputation really did help generate tens of billions of dollars in revenue in a remarkably short time.

But a strategy is only as strong as its ability to survive pressure, and 2026 supplied real pressure: rivals racing ahead on defense contracts, competitors moving faster with fewer public safety commitments, and a market that rewards whoever ships next, not whoever is most careful. Anthropic's decision to loosen part of its own safety framework under exactly that pressure doesn't prove the strategy was hollow. It proves the strategy has a cost, and that cost is now being tested in public.

Daniela Amodei didn't just help build an AI safety company. She helped prove that safety could be sold, and now she's finding out what happens to that pitch when the competition stops playing by the same rules.

Sources & References

• The Long-Term Benefit Trust, Anthropic (2023)
• Anthropic's Responsible Scaling Policy, Anthropic (2026)
• Anthropic Overhauls Core Tenet of Its Safety Policy, PYMNTS (2026)
• Anthropic says it hit a $30 billion revenue run rate after 'crazy' 80x growth, VentureBeat (2026)
• Dario Amodei and Daniela Amodei, Time (2026)