Anthropic reportedly views itself as the antidote to OpenAI's "tobacco industry" approach to AI
That OpenAI ended up creating its most dangerous rival, Anthropic, could one day be one of the defining stories of the tech industry.
In a detailed report, Sam Altman biographer Keach Hagey digs into the history of the split. The official narrative is that OpenAI wasn't taking safety seriously enough for the team that would go on to found Anthropic, and according to Hagey, there's plenty of evidence to support that version of events.
But her account also reveals that the breakup ran deeper than ideology: personal slights, power struggles, and a constant feeling of being overlooked, especially in the relationship between Greg Brockman and Dario Amodei.
According to the report, early conflicts included a debate over whether AGI could one day be sold to governments or the nuclear powers on the UN Security Council. Amodei found the idea completely unacceptable. Brockman did not.
The dispute later escalated over control of the language model projects. Amodei was one of the key figures behind GPT-2 and GPT-3 and apparently tried to keep Brockman's influence over that work to a minimum. At the same time, according to Hagey, he repeatedly felt sidelined, whether it was visibility, recognition, or access to important meetings, like being left out of a meeting with Barack Obama. By the end of 2020, Dario and Daniela Amodei left the company with several other OpenAI employees to start Anthropic.
| Time | Event | Significance |
|---|---|---|
| 2016 | Dario Amodei joins OpenAI; early philosophical differences with Greg Brockman emerge. | The fundamental conflict begins: safety and caution vs. growth and aggressive development. |
| 2017-2018 | After internal crises and departures (partly tied to Elon Musk), Amodei loses trust in parts of OpenAI's leadership. | Disagreements turn into a leadership and trust crisis. |
| 2018-2019 | Dispute over control and influence in the language model projects. | The conflict shifts to power, responsibilities, and recognition. |
| 2019-2020 | According to the report, Amodei repeatedly feels sidelined; management tensions escalate. | Personal grievances and status issues deepen the rift. |
| End of 2020 | Dario and Daniela Amodei and other employees decide to leave. | The internal conflict becomes irreparable. |
| 2021 | Anthropic is founded. | A direct competitor emerges from OpenAI's internal crisis. |
| To this day | OpenAI and Anthropic compete over market share, talent, and the AGI narrative. | The original split continues to shape the AI industry. |
The rivalry still shapes the AI industry today
The conflict between OpenAI and Anthropic continues to play out in the open as competition for talent, capital, and customers, but more than anything as a fight over how A(G)I should be developed and brought to market.
According to Hagey, Anthropic internally sees itself as a "healthier alternative" to OpenAI, with people at the company comparing Altman's operation to the tobacco industry. When OpenAI picked up a Pentagon deal that Anthropic had walked away from, Amodei reportedly called Altman "mendacious" internally. According to the report, he wrote that the events pointed to "a pattern of behavior that I've seen often from Sam Altman."
Amodei also called Greg Brockman's $25 million donation to a pro-Trump Super PAC "evil," though his own company has done its share of cozying up to the Trump administration—although the Pentagon fallout made it pretty clear where Anthropic actually stands.
Even during Altman's brief removal as CEO, there were internal accusations that he had been manipulative and dishonest. After his return, however, an internal investigation cleared him of those charges.
AI News Without the Hype – Curated by Humans
As a THE DECODER subscriber, you get ad-free reading, our weekly AI newsletter, the exclusive "AI Radar" Frontier Report 6× per year, access to comments, and our complete archive.
Subscribe now