Daniela Amodei: Leaving OpenAI Was 'Kind of a Crazy Thing to Do'

Anthropic's president didn't walk away from OpenAI with confidence. She walked away with doubt, months of internal debate behind her, and the quiet push of a trusted mentor who chose not to tell her what to do. In a conversation reported by Benzinga this week, Daniela Amodei reflected on the late-2020 departure that produced one of the most consequential bets in the history of artificial intelligence, describing it in terms that don't fit the triumphant founding narrative the company's current $965 billion valuation might suggest.

The Decision Before the Decision

Amodei left OpenAI in late 2020, along with her brother Dario Amodei and a group of senior researchers including Tom Brown, Chris Olah, Sam McCandlish, Jack Clark, and Jared Kaplan. The team had spent years near the center of what was then the most ambitious AI research organization in the world. By the time they left, they believed the company's direction had shifted in ways that made staying an implicit endorsement of an approach they couldn't fully support.

The departure wasn't clean. The group spent months debating whether the disagreements were fundamental or manageable, whether they could still make things work from inside, and whether the timing was right. At one point Daniela Amodei called a trusted friend and mentor outside the company to work through it. The response surprised her in its brevity. The mentor told her: "Honestly, I don't think you really need to be on the phone with me. You already know what the right answer is."

She has since described leaving OpenAI as "kind of a crazy thing to do." At the time, OpenAI was the most prestigious address in AI research. The people who eventually formed Anthropic were turning down the option to keep building on top of GPT-3 and its successors, at a company that had just secured a $1 billion commitment from Microsoft. The uncertainty was real.

Safety as Business Strategy

Anthropic was built on the premise that safety and commercial success don't have to conflict. Amodei has made this case in multiple settings over the past year, including a View From the Top appearance at Stanford Graduate School of Business in May. Her position is that most enterprises are not looking for unsafe models. A company with a serious track record on safety and governance can move faster in regulated markets than a competitor that treats safety as a cost center.

The argument has been tested against reality at a scale the founders probably didn't anticipate. Anthropic closed a $65 billion Series H round in May, reaching a $965 billion valuation that overtook OpenAI's. Revenue is running at an annualized rate of $47 billion, driven by enterprise deployments of Claude across financial services, healthcare, legal, and software development. Daniela Amodei's own stake, according to Forbes, is now worth approximately $7 billion.

"Honestly, I don't think you really need to be on the phone with me. You already know what the right answer is." Daniela Amodei's mentor, as recounted to Benzinga, June 2026

What the IPO Chapter Means

Anthropic filed a confidential IPO prospectus with the SEC in June 2026, moving toward a public offering that would give retail investors their first chance to own a share of the company. The filing marks a clear point of resolution to a founding story that was, by Amodei's own telling, far less certain than the outcome makes it look.

The arc from "kind of a crazy thing to do" to the most valuable AI startup in the world carries a specific lesson for the people who work in and around AI today. It isn't that safety-first AI was always obviously going to win commercially. It's that the people who built Anthropic believed it would, accepted the risk that it might not, and structured everything around that bet rather than hedging it.

For context on how Anthropic has structured its approach to AI values and training, the company's Constitutional AI documentation lays out the technical underpinning. Daniela Amodei's Stanford remarks from May cover the same ground from a business strategy angle. The IPO filing, when it becomes public, will show whether the market prices the safety-first positioning as an asset or a cost.

For now, the founding story has its answer. Five years after a group of researchers made what Amodei calls a crazy bet, the bet is the biggest in the industry. Learn more about Anthropic's history and mission for full context on how the company got here.