SSI Raises $1 Billion to Research Safe Superintelligence

Safe Superintelligence (SSI), Ilya Sutskever’s new company, has raised $1 billion to research artificial intelligence systems more capable than today’s models. The deal values the company at $5 billion, even though it has yet to release a product or explain what model it is developing.

The figure confirms that investors remain willing to fund very early-stage AI projects when they are led by researchers with significant standing in the field. Sutskever was a co-founder and chief scientist at OpenAI, where he helped develop some of the most influential language models of recent years.

A company focused on a single problem

SSI was introduced in June by Sutskever alongside Daniel Gross, Apple’s former AI chief, and Daniel Levy, a researcher who also worked at OpenAI. Its approach is unusual even for a frontier startup: the company says it will not divide its efforts between commercial products and safety research.

Its goal is to create safe superintelligence, a system that vastly exceeds human cognitive abilities without acting against people’s interests. Superintelligence does not exist today as a proven technology. It is a hypothesis about a future stage of AI, beyond today’s models that generate text, images, code or answers to questions.

In this context, safety is not limited to preventing harmful responses from a chatbot. The underlying challenge is how to design, evaluate and control systems that could solve complex tasks with a high degree of autonomy. It is an open question: OpenAI, Google DeepMind and Anthropic have not demonstrated that they know how to guarantee that level of control for AI with capabilities far beyond those of humans.

$1 billion with no application in sight

The round included firms such as Andreessen Horowitz, Sequoia Capital, DST Global and SV Angel, as well as NFDG, the fund run by Daniel Gross and Nat Friedman. The $5 billion valuation puts SSI among the sector’s most expensive private bets at such an early stage.

This is not an investment based on revenue, users or a product that can be compared with ChatGPT, Claude or Gemini. It rests on the founding team and the conviction that the next generation of models will require more research, greater computing power and new safety techniques.

That explains both the size of the investment and the risk. Training advanced models requires data centers, specialized chips and research teams that are difficult to hire. But a high valuation before showing results also raises expectations: SSI will have to prove that its approach produces technical advances, not just an appealing thesis about AI risks.

The market rewards frontier teams

The funding arrives just months after Sutskever left OpenAI. His departure followed the governance crisis that engulfed the company in November 2023, when its board briefly ousted Sam Altman before he returned as chief executive. Sutskever ultimately left OpenAI in May.

SSI is part of a new wave of labs trying to compete at the most expensive layer of artificial intelligence: building foundation models. These systems later serve as the basis for assistants, enterprise tools, code generators and scientific applications.

For companies that use AI, the round does not immediately change which tools they can deploy. SSI has not announced an API, an assistant or launch dates. Its significance lies elsewhere: it shows that advanced-model safety has become a business bet in its own right, rather than merely an internal focus for the major labs.

The next indicator will be technical. The company will have to spell out how it plans to research alignment—the set of methods intended to make AI follow human goals—and whether it can attract enough talent and computing capacity to turn that ambition into verifiable results.