Sutskever founds Safe Superintelligence

Ilya Sutskever, OpenAI’s former chief scientist, announced Safe Superintelligence Inc. (SSI) on X, roughly a month after leaving OpenAI’s board and the company he had co-founded, and days after Jan Leike resigned as head of the Superalignment team he and Sutskever had led together. He co-founded SSI with Daniel Gross, former head of AI at Apple, and Daniel Levy, an OpenAI alumnus and investor; the company split its operations between Palo Alto, California, and Tel Aviv, Israel.

SSI described itself as pursuing “one goal and one product: a safe superintelligence,” framing safety and capability advancement not as a trade-off to be balanced against product deadlines but as a single technical problem to be solved together, so that the company could “scale in peace.” Its published materials made no commitment to release any product before reaching that goal, and reporting on the launch described it as a pure research organisation with no near-term plan to sell AI products or generate revenue — a structure explicitly designed to insulate research from the commercial and release pressures Sutskever and Leike had both cited as reasons for OpenAI’s Superalignment team losing out on compute and priority.

The founding continued a wave of departures from OpenAI’s safety-focused ranks that followed the November 2023 board crisis, in which Sutskever had briefly voted to remove Sam Altman before publicly regretting it. SSI’s launch was read at the time as a direct statement about that dispute: rather than trying to reform OpenAI’s balance of safety and product from within or from a rival’s product team, Sutskever chose to found a company structured, from its first announcement, to make that trade-off structurally impossible. The company went on to raise substantial funding on the strength of Sutskever’s reputation alone, despite shipping no public product in its first year.