OpenAI co-founder Ilya Sutskever has been named CEO of his artificial intelligence startup Safe Superintelligence. His predecessor, Daniel Gross, recently left, reportedly to join Meta’s new superintelligence division dedicated to building AI models that could surpass human capabilities.
Sutskever told his team that Gross’s work at Safe Superintelligence had been “winding down,” something they were already aware of, in a message he later posted on X. He also said that the company’s co-founder, Daniel Levy, would be promoted from Principal Scientist to President.
Last week, in an internal memo, Mark Zuckerberg revealed the team members of Meta Superintelligence Labs. It included Alexandr Wang, the founder of data-labelling startup Scale AI, former GitHub CEO Nat Friedman, and 11 engineers from the likes of OpenAI, Google DeepMind, and Anthropic. The Meta boss had reportedly been offering $100 million signing bonuses to score the team he wanted, which will be led by Wang.
Gross was conspicuously absent from the memo. Zuckerberg approached Gross after Sutskever declined his proposal to join Meta and sell Safe Superintelligence earlier this year, sources told CNBC. The company was valued at over $30 billion in February, despite not having a product or revenue.
Sutskever hinted at this in his CEO announcement. “You might have heard rumors of companies looking to acquire us,” he wrote on X. “We are flattered by their attention but are focused on seeing our work through.”
While Meta has yet to confirm Gross’s hiring, there are signs of a budding partnership. This week, Facebook’s parent company reportedly offered to buy a minority stake in NFDG, the venture firm Gross co-founded with Nat Friedman, according to the Wall Street Journal. If both founders are now at Meta and NFDG is no longer making new investments, as Friedman confirmed, Meta’s offer could give limited partners a chance to cash out at full value.
The race to superintelligence continues
Sutskever, who left OpenAI in May 2024 after nearly a decade as its chief scientist, co-founded Safe Superintelligence in June 2024 with Gross, a former AI lead at Apple, and Levy, an ex-OpenAI researcher. Now, as CEO, Sutskever says the technical team will continue reporting to him, emphasising that the company does not need outside interference to achieve its mission.
“We have the compute, we have the team, and we know what to do,” Sutskever said in his X post. “Together we will keep building safe superintelligence.”
Nevertheless, he is not alone in chasing that goal. When Zuckerberg announced Meta Superintelligence Labs, he said his company was “uniquely positioned” to deliver superintelligence. He said Meta has enough cash to acquire the significant compute resources required, a track record of delivering products to billions of people, experience in AI hardware such as smart glasses, and a company structure that allows it to take big swings.
Whether safety will be a priority, like it is at Safe Superintelligence, remains to be seen, but Meta’s reputation regarding privacy and security has not been stellar.
Meanwhile, OpenAI announced in January that its primary goal for the year would be the development of superintelligence. Then, last month, Sam Altman said his company had “recently built systems that are smarter than people in many ways.”
Sam Altman thinks that Meta poaching OpenAI employees for Meta’s new superintelligence division is “somewhat distasteful.” Find out why this is somewhat ironic here.