Before OpenAI became one of the most prominent companies in artificial intelligence, its early years reportedly included some highly unusual business proposals. According to a new account of the company's founding era, leaders floated selling future AI tools to governments that included China and Russia, and there was also a plan to give Kanye West early access to an unreleased model. Top staffers reportedly reacted with alarm, and the objections they raised at the time continue to resonate across the AI industry today.

The account, reported by technology journalist Kevin Roose in The Atlantic, carries weight because it helps explain one of the most consequential splits in modern tech: several of the people who pushed back hardest on these ideas later walked out and founded Anthropic, now one of OpenAI's most direct rivals. It also lands amid a debate that has only intensified as AI has become big business: who should fund and shape technology this powerful, and who gets to use it.

The 'Countries Plan' and the Kanye West Pitch

The central episode dates to 2017. OpenAI co-founders Greg Brockman, now the company's president, and Ilya Sutskever reportedly advanced an internal proposal known as the 'countries plan.' The idea was to fund the race toward artificial general intelligence, or AGI, by bringing nation-states into the picture. Under the proposal, OpenAI could have auctioned off future AGI rights or accepted money from governments. The countries mentioned included the United States, China, and Russia. Who funds frontier AI has since become a live policy question, as the cost of training cutting-edge models has climbed to levels few organizations can shoulder alone.

Dario Amodei, then a senior researcher at OpenAI, reportedly objected strongly. In September 2017, he laid out his concerns and called the plan "morally illegitimate." His reasoning, according to the report, was twofold: he argued the plan could intensify arms races between rival powers and could help enable human rights abuses. Both concerns have since moved into mainstream policy debates, from government controls on advanced AI chips to international discussions of AI safety and military use.

The second episode came in 2020. Sam Altman, OpenAI's chief executive, reportedly offered Kanye West early access to a version of GPT-3, the language model that would later become a foundation for OpenAI's commercial products. That same year, OpenAI began offering GPT-3 to outside developers through a paid API, an early step in its commercial turn. OpenAI's policy director reportedly shut the arrangement down, pointing to West's unpredictable behavior.

OpenAI has pushed back on parts of the story. The company disputes the characterization of the countries plan, saying it only considered an international cooperative model and never intended to sell AGI to China or Russia.

How Internal Fights Became a Rival Company

Amodei was not a peripheral figure at OpenAI. He played a key role in developing GPT-2 and GPT-3, the models that turned the company from a research curiosity into a serious commercial player.

According to the report, the countries plan and the Kanye West episode were part of a wider pattern of friction. Disagreements over governance, the push toward commercialization, and trust in leadership all reportedly piled up during OpenAI's shift toward a hybrid commercial model. Founded in 2015 as a nonprofit focused on building safe AI for broad benefit, the organization moved toward a structure that could raise large sums and sell products.

Amodei and several colleagues eventually left. In 2021, they founded Anthropic, pitching it as a laboratory that would put safety first, with Amodei serving as chief executive.

What This Means for the AI Race

The rivalry between OpenAI and Anthropic reportedly began as a disagreement over values: who should fund powerful AI, who should get access to it, and what guardrails belong around both. Both labs now compete for enterprise customers, government contracts, and top research talent. Their origin story continues to shape how each one presents itself, with Anthropic leaning heavily on a safety-first identity that this account helps explain.

OpenAI's denial will be central to how the episode is received. If the company's version holds, the countries plan looks like an early brainstorm about international cooperation that never went anywhere. If critics' version gains traction, it raises harder questions about judgment at the top of a company that now wields enormous influence. Any additional on-the-record accounts from former staff, or a fuller company rebuttal, will help determine which version sticks.

The Kanye West episode, meanwhile, illustrates how informal access decisions could be made inside a company handling powerful technology—and how much depended on individual staff willing to say no.