Safe Superintelligence (SSI)

in English: Safe Superintelligence Inc. (SSI)

basicexperimental

An American AI research company with offices in Palo Alto and Tel Aviv, created for a single purpose: to build artificial intelligence smarter than humans but safe, without intermediate products.

In plain terms

Safe Superintelligence is an AI lab founded by Ilya Sutskever, one of the founders and former chief scientist of OpenAI, along with two partners. The company has a single goal: to build artificial intelligence that surpasses humans in every intellectual task while remaining reliably safe.

Imagine a car manufacturer announcing that it will not release a new model every year, hold no showcases, and sell nothing until it has built the one car that never causes an accident. SSI works the same way: it has no chatbot, no subscription, and no public model.

This differs from OpenAI, Google, or Anthropic, which release products continuously and also fund research from their revenue. According to SSI, product competition and short-term business pressure distract from safety. That is why the company does its research behind closed doors, funded by investors.

How it works

The company operates with a small, select research team at two locations: Palo Alto, California, and Tel Aviv.
Operations are funded by investors' money: venture capital funds and large technology companies, such as Alphabet and Nvidia. There is no publicly known revenue because it sells nothing.
Most of the money goes to computing capacity, i.e., the machines on which large AI models are trained.
Security and capabilities are treated together as a single technical task: capabilities are to be increased as fast as possible, but with security always staying one step ahead.
Research results are barely made public, so it is impossible to judge from the outside where they stand.

An everyday example

You read in economic news that a company received a valuation of tens of billions of dollars despite having neither products nor revenue. This is SSI: investors are betting on the founder's reputation and a technology that does not yet exist. You will not encounter its app because it does not exist.

Who is affected

bank customer, investorSSI's valuation clearly shows how large sums move in the AI market even without a product or revenue, which also raises the risk of a bubble. → Money

developerThe company's stance influences the professional debate over whether security should be bolted onto models after the fact or designed in from the start. → Code

lawyer, citizenIf a closed private company builds the most powerful AI, the question arises: who oversees it and who is responsible? → Law

everyoneIf superintelligence is ever built, it will affect everyone's life, so it matters how safely it is constructed.

A common misconception

MythSSI has already built the safe superintelligence, it just has not released it yet.

In realityThere is no public evidence for this. The company's name describes its goal, not a finished product, and no one has yet demonstrated superintelligence.

Limits and risks

The company shares little about its work, so its claims cannot be verified externally.
The word 'secure' in the name is a promise that has not yet been verified by independent investigation.
It is also uncertain whether superintelligence can be built at all, and if so, when: researchers' opinions on this are strongly divided.
Investors will eventually expect returns, which could pressure the 'no product' principle. Sutskever himself signaled in late 2025 that this plan may change over time.

Similar, but not the same

Szuperintelligencia – Superintelligence itself is the hypothetical AI smarter than humans, while SSI is a company aiming to build it.
OpenAI – OpenAI releases products and generates revenue from them, while SSI, according to its announced plan, releases nothing until it reaches its final goal.
AI safety – AI safety is an entire research field that many organizations work on; SSI is just one of the players.

Related concepts

Broader

AI-labor, AI safety, Artificial intelligence

Narrower

–

On the edge

Szuperintelligencia, Artificial general intelligence (AGI), OpenAI, Anthropic, Alignment

A brief history

The company was founded in June 2024 by Ilya Sutskever, Daniel Gross, and Daniel Levy, one month after Sutskever left OpenAI. In autumn 2024, it raised 1 billion dollars, and in spring 2025, an additional 2 billion dollars, the latter reportedly at a valuation of approximately 32 billion dollars. In summer 2025, Gross moved to Meta; since then, Sutskever has been the CEO and Levy the president.

Where things stand

According to the last public information, SSI is in a closed research phase: it has raised billions in capital but has no product and has not presented a model. Nothing is known about its results.

Status: 2026 November

Also known as

SSI · Safe Superintelligence · Safe Superintelligence Inc. · SSI Inc. · ssi.inc · Safe Superintelligence · Sutskever's company · Ilya Sutskever startupja

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.