OpenAI CEO Sam Altman reportedly tried to deceive the US government into giving his company billions of dollars by warning of an AI race that may not have actually existed. In 2017, Altman met with Us intelligence officials and claimed that China had launched an ‘AGI Manhattan Project.’ He urged that his company needs billions of dollars of government funding to keep pace. But when officials pressed him for evidence, Altman said, ‘I’ve heard things.’ He made similar claims in multiple meetings and promised to provide proof later, but he never did.
After looking into the matter, officials found no evidence that such a Chinese project existed, reports The New Yorker. One intelligence official later said, ‘It was just being used as a sales pitch.’ This suggests that Altman may have exaggerated or even invented the threat to push the government into funding OpenAI.
Altman often compared the development of AGI to the Manhattan Project, but he adjusted the meaning depending on the audience. When speaking to government officials, he stressed urgency and competition, suggesting that the US needed to act fast and invest heavily. However, when talking to safety-conscious audiences, he warned about the dangers of AGI. and argued for careful, internationally coordinated development, as per the report.
Also read: Your favourite apps may be tracking you: Here is how to stay safe
Inside OpenAI, there were also disagreements about how to move forward. Page Hedley, a former policy and ethics adviser, suggested creating global partnerships between AI labs to prevent a ‘catastrophic’ arms race. However, other company members prioritised growth and competition. According to several interviews and contemporaneous records, leadership often pushed back on ideas.
One proposal reportedly involved using global rivalries to OpenAI’s advantage. The idea was to encourage countries like China and Russia to compete for access to OpenAI technology, possibly even starting a bidding war. According to Hedley, the thinking was simple: It worked for nuclear weapons, so why not for AI?
If the report is true, it raises serious questions about how far tech leaders are willing to go to gain government support.