Ilya Sutskever launches Safe Superintelligence Inc.

A month after officially leaving OpenAI, Ilya Sutskever is back with a new venture. And true to form, he is not building an AI search engine or another enterprise chatbot.

He is launching Safe Superintelligence Inc., or SSI, an AI lab created with a singular focus: build safe superintelligence, and do literally nothing else until it is done.

One product, zero distractions

SSI is not positioning itself as a standard startup. According to its launch statement, the company has one goal and one product, approaching safety and capabilities in tandem.

Sutskever is teaming up with two notable co-founders: Daniel Gross, a former AI lead at Apple, and Daniel Levy, who previously worked as a technical staff member at OpenAI.

What stands out most is the direct shot at how Big Tech operates. The announcement emphasizes that SSI's singular mission insulates it from management overhead, short-term commercial pressures, and endless product release cycles.

As Sutskever put it in an interview with Bloomberg, the company will not release a single commercial tool until safe superintelligence exists. The goal is to simply scale in peace.

The fallout from OpenAI becomes a new battle line

This launch puts the dramatic fracture at OpenAI into sharp focus.

Sutskever led the board revolt that briefly ousted Sam Altman last year. While Altman reclaimed power within days, Sutskever eventually stepped down from the company in May, hinting at a new project.

He was not alone. Shortly after his departure, alignment researcher Jan Leike resigned, publicly stating that safety culture and processes at OpenAI had taken a backseat to shiny products. Policy researcher Gretchen Krueger exited around the same time with similar safety concerns.

While OpenAI races to cement massive distribution deals with Apple and Microsoft, Sutskever is running in the exact opposite direction. He is trading product velocity for scientific isolation.

Why it matters

For everyday users and developers, do not expect an API key, a playground, or a consumer app anytime soon. SSI is deliberately not building for you today, and they are entirely upfront about it.

For the AI landscape, this is a radical counter-experiment against Silicon Valley's relentless commercialization treadmill. While OpenAI and Google scramble to monetize every monthly update, Sutskever is betting that real safety breakthroughs can only happen without quarterly revenue pressures.

We think this is a brilliant recruitment pitch for world-class researchers who are genuinely exhausted by ship-it-yesterday product cycles. If you want to work on foundational intelligence without worrying about enterprise sales quotas, SSI looks like the ultimate research sanctuary.

However, the strategy carries a massive financial question mark. Training frontier models demands billions of dollars in compute, which is traditionally subsidized by enterprise revenue and massive commercial partners. Whether investors will fund an open-ended research effort that explicitly refuses to generate intermediate revenue is the biggest gamble of Sutskever's career.