OpenAI Explored Partnerships With Russia, China and Gave Kanye West Early Access to AI Model, Report Says

OpenAI considered drastic proposals, including offering future artificial intelligence tools to Russia and China and providing early access to an unreleased model to Kanye West, according to a report.

These ideas, detailed in The Atlantic, reportedly alarmed top staff and contributed to the departure of Dario Amodei, then OpenAI's Vice President of Research. Amodei subsequently founded Anthropic, a company focused on AI safety.

One plan, dubbed the "countries plan," was reportedly conceived in 2017 by OpenAI co-founders Greg Brockman and Ilya Sutskever. The proposal aimed to raise funds by auctioning rights to future artificial general intelligence (AGI) technology to governments, including those of China and Russia. A source described the concept as "insane," likening it to tech entrepreneurs pitching a "wacky plan" to make governments fund it.

Amodei expressed strong objections to the "countries plan" in an email to OpenAI executives. He argued that such a deal would escalate arms races, potentially enable human rights abuses, and damage OpenAI's moral standing. He specifically raised concerns about granting advanced AI capabilities to leaders like Vladimir Putin, warning of the "unspeakably awful things" that could result.

This plan did not proceed, reportedly due to OpenAI's subsequent substantial funding from Microsoft.

Another incident involved CEO Sam Altman offering Kanye West access to a pre-release version of GPT-3 in 2020. OpenAI's policy director, Jack Clark, reportedly blocked the offer due to West's increasingly erratic public behavior, including antisemitic remarks and a video that appeared to show him urinating on an award. Clark noted West's admiration for Adolf Hitler shortly after the decision.

Amodei also reportedly felt that Altman was disingenuous in communications regarding negotiations with Microsoft. He suspected Altman did not fully disclose safety provisions to Microsoft while also telling Amodei that Microsoft had rejected them.

These experiences, among others, led Amodei to secretly recruit seven other OpenAI employees, including Jack Clark and his sister Daniela Amodei, to establish Anthropic in 2021.

An OpenAI spokesperson told The Atlantic that the company "briefly considered whether to turn OpenAI into an international cooperative endeavor" but denied considering the sale of AGI to China or Russia.