CHAPTER 1A different question at the start
Most startup stories begin with an unmet customer need. Anthropic's begins with an argument about how to build powerful AI responsibly. Dario and Daniela Amodei and other researchers left OpenAI and started Anthropic in 2021. It was a new lab in a field where every new model demanded costly talent and computing power. Reuters' timeline records a $124 million Series A in May 2021, $580 million Series B in April 2022, and a Claude model released in March 2023. That model turned an abstract research promise into a service a customer could actually test.
It would be too neat to turn the departure into a simple tale of good people leaving a bad company. The public accounts establish the founding, the founders' prior work and safety as a central concern. They do not disclose every person's private motive. Nor can a founding promise alone prove every later product choice was safe. The real question is whether the company can earn trust while making models useful enough for people to buy.
For a retailer, the issue is recognizable. A product must do what the label says. With AI, the label is harder to check: a model can appear capable on ordinary tasks and fail under a rare, costly condition. Businesses need reliability, controls and a path to human review as much as they need a clever answer. Anthropic built an identity around taking those questions seriously. Whether it can keep that identity under the pressure of a race to ship is part of this story.
Sources: Reuters company chronology, https://www.reuters.com/technology/anthropics-path-ai-startup-industry-defining-ipo-2026-09-28/ ; Reuters IPO report, https://www.reuters.com/business/ai-giant-anthropic-confidentially-files-us-ipo-2026-06-01/ .