What is Safe Superintelligence (SSI), and is it the same as SI? (2026)
What is Safe Superintelligence? SSI is the AI lab Ilya Sutskever co-founded in June 2024 to build one product, a safe superintelligence. As of October 2026 it has raised billions, shipped nothing, and is not the same as "SI", the US government's new name for ordinary AI.
Safe Superintelligence Inc. (SSI) is the AI research lab Ilya Sutskever co-founded in June 2024 after leaving OpenAI. It has one stated goal, a safe superintelligence, and as of October 2026 it has shipped no model or product. It is not the same thing as "SI", the label a September 2026 US executive order now uses for ordinary AI.
That last part is why the question got harder in the past three weeks. Before 29 September, "what is safe superintelligence" had one answer. Now an assistant has to tell apart a company called SSI, a government term called SI, a sci-fi idea called ASI and a Slovenian domain that ends in .si. And most of the pages that rank for it were written before the rename, so they still treat the August model rumours as if they were live news.
Below: what SSI is, who runs it, what it has raised, what it has really put out, and why the name clash matters if you build an AI product. Every figure links to the page it came from. Where a number is only reported, we say so.
What is Safe Superintelligence Inc. (SSI)?
SSI is an American AI lab with offices in Palo Alto and Tel Aviv. Its own homepage describes it as a company with one goal and one product: a safe superintelligence. It says it treats safety and capability as technical problems to solve together, and that its structure keeps it away from short-term commercial pressure.
Put simply, it is trying to skip the step every other frontier lab took. OpenAI, Anthropic and Google pay for research by selling chat apps and APIs. Sutskever's team has raised billions without selling a thing, on the bet that the first thing it ships will be the one it set out to build.
"Superintelligence" in SSI's name carries the older, research meaning: a system that outperforms the best humans across most intellectual work. That is a future capability, not a product category. We cover how close forecasters think that is in What is superintelligence, and how close are we?
Is SSI the same as SI, the new name for AI?
No. SSI is a company. SI is a word swap. The executive order Inaugurating the Era of Super Intelligence, signed on 29 September 2026, tells federal agencies to write "Super Intelligence" and "SI" where they used to write "artificial intelligence" and "AI". It defines SI as the same technologies the existing statutory definition of AI already covers, under 15 U.S.C. § 9401(3).
So under the order, a spam filter is SI. By SSI's meaning, superintelligence doesn't exist yet. Same word, opposite ends of the scale.
The order applies to the executive branch only. It places no duty on private companies, and it gives the President's science adviser 60 days, to about 28 November 2026, to propose a statutory definition. Some companies are adopting the label anyway. On 9 October, Salesforce CEO Marc Benioff posted that "AIForce is officially SIForce", as reported by TheStreet. Every rename like that nudges "SI" a bit closer to meaning "the AI we already have", and pushes the lab's sense of the word further back.
Here is how the five names line up as of October 2026:
| Name | What it is | Who uses it | Exists today? |
|---|---|---|---|
| SSI (Safe Superintelligence Inc.) | A private research lab | Ilya Sutskever and team | The company, yes. The product, no |
| SI ("super intelligence") | A new federal term for ordinary AI | US executive branch, some companies | Yes, it is today's AI |
| ASI (artificial superintelligence) | A hypothetical system smarter than humans at most things | Researchers, forecasters | No |
| Meta Superintelligence Labs | Meta's umbrella AI division, formed in 2025 | Meta | Yes, as an org chart |
| .si | Slovenia's country-code domain | Anyone who registers one | Yes |
If you want the long version of the SI, AGI and ASI split, read SI vs AGI vs ASI: what's the difference?. For the order itself, our guide What is SI (super intelligence)? walks through it section by section.
Who founded SSI, and who runs it now?
Ilya Sutskever runs SSI as CEO, with Daniel Levy as President. The company was announced on 19 June 2024 by three co-founders: Sutskever, Daniel Gross and Levy, according to Wikipedia's SSI entry. Gross left in mid-2025, and the leadership changed around his exit.
Sutskever co-founded OpenAI and was its chief scientist. TIME's 2026 AI profile credits him with overseeing the research behind ChatGPT and the early work that led to o1. He left OpenAI in May 2024, months after his role in the brief removal of Sam Altman as CEO in November 2023.
Gross had previously led AI efforts at Apple. Levy is a former OpenAI researcher. SSI's updates page records that Gross left on 29 June 2025, that Sutskever became CEO and Levy became President, and that the company turned away acquisition interest to keep working on its own plan. Gross went on to join Meta Superintelligence Labs, per the same Wikipedia entry. That is the division Mark Zuckerberg announced on 30 June 2025 under Alexandr Wang and Nat Friedman, according to Fortune.
For a lab with that much cash, the team is tiny. Globes reported in August 2026 that SSI had hired just over 50 people across its two offices.
How much money has SSI raised?
SSI has raised somewhere between $3 billion and $7 billion, depending on whether you count Nvidia's 2026 investment, whose size neither company has confirmed. The figures that come from SSI itself are thinner than most coverage suggests, so here is each one with its status:
| Date | Event | Amount | Status |
|---|---|---|---|
| 4 Sep 2024 | First round (NFDG, a16z, Sequoia, DST Global, SV Angel) | $1B, at a reported $5B valuation | Amount confirmed by SSI |
| 2025 | Second round | $2B at a $32B valuation | Reported by TechCrunch |
| Apr 2025 | Google Cloud deal for TPU compute | Not disclosed | Reported, per Wikipedia |
| 26–27 Jul 2026 | Nvidia strategic partnership, access to Vera Rubin | Undisclosed. Bloomberg reported $5B | Partnership confirmed by Nvidia. Amount unconfirmed |
| Aug 2026 | Running total | About $7B over two years | Reported by Globes |
Two things jump out. Nvidia's own press release mentions an investment but gives no number, so any page that states "$5 billion" as fact is leaning on Bloomberg's sources. And SSI's July update says the deal lets it scale compute about 10x, which Nvidia phrases as an order of magnitude more on Vera Rubin. Sutskever's line in that release: "We have research that is worthy of scaling up."
Has SSI released a model or product yet?
No. As of 26 October 2026, SSI has released no model, API, paper, waitlist or product. Its updates page has three entries in total: the 2024 funding round, the 2025 leadership change and the July 2026 Nvidia partnership. Nothing after that.
Worth saying plainly, because a lot of August coverage hinted otherwise. On a podcast in early August, investor Gavin Baker said SSI would release a model that month. Globes relayed his claim on 26 August, alongside unnamed experts who said a model was being tested, and reported that the rumoured system updates its own weights as it reads documents. Sutskever did not comment. August ended. Nothing showed up on any official SSI channel.
That quiet fits how Sutskever behaves in public. TIME notes he rarely talks about direction, though he has pointed at two limits of today's systems: they stop learning once training ends, and they need far more examples than a person to pick up something simple. A model that keeps learning after it's deployed would fix the first. Does SSI have one? Nobody outside the company knows, and anyone who says they do is guessing.
How is SSI different from OpenAI, Anthropic and Meta Superintelligence Labs?
Mostly, it's revenue. OpenAI, Anthropic and Meta all ship products to millions of people and pay for part of their research that way. SSI ships nothing, on purpose. There's no chat app, no sales team and no API pricing page.
The practical differences for someone choosing tools today:
- You cannot use SSI. There is no product to sign up for, so it has no place in your stack, your vendor list or your comparison pages.
- You cannot benchmark it. With no model card or evals, any claim about SSI's capability is speculation.
- It still shapes the market. Nvidia backing a lab with no product, and handing it Vera Rubin compute, tells you where serious money thinks the next jump comes from.
- Meta's lab is the opposite bet. Meta Superintelligence Labs sits over Meta's existing product, research and FAIR teams, right next to products Meta already ships. Same word in the name, very different plan.
Why does the SSI vs SI confusion matter if you run an AI startup?
Because AI assistants now have to work out what "SI" means every time someone asks, and they lean on whichever pages explain it most clearly. If your site says "SI" with no context, a model has to guess which of the five meanings you had in mind.
Say you run a small AI meeting-notes tool and you changed your homepage hero from "AI-powered" to "SI-powered" the week after the order. A founder asks Perplexity, "What SI tools take meeting notes?" Half of what the assistant pulls back is about Sutskever's lab and the theory of superintelligence. Next to that, your page either reads like a claim to be far more than a notes app, or it gets skipped because it doesn't fit. Neither gets you cited.
The fix is cheap: keep "AI" on the page, add "SI" next to it once, and say what you mean. We made the full case in Should I rebrand my AI startup to SI? The short version: the order binds federal agencies, not you, and your customers still type "AI".
One more reason to care: whoever is asking assistants about SI this month is an early adopter almost by definition. When they land on your site, you'll want to know which assistant sent them and what they read.
Do this tonight
- Ask ChatGPT, Perplexity and Claude "what is SI?" and "what is SSI?". Note whether they mix the two up, and which pages they cite.
- Search your own site for every use of "SI", "superintelligence" and "super intelligence". Make sure each one says which meaning it intends.
- If you added "SI" to your homepage after 29 September, keep "AI" in the title tag and the first heading too.
- Bookmark ssi.inc/updates. It is the only source that counts when the next "SSI model is coming" rumour lands.
- Check whether AI crawlers can read your key pages, so assistants quote your definition rather than someone else's.
If you want to see which assistants actually send you visitors, Clerion reports ChatGPT, Perplexity, Claude, Gemini and Copilot as their own channel, and the free AI crawler checker shows what a model sees on any page.
Related reading
- What is SI (super intelligence)? The new name for AI, explained: the executive order, line by line.
- SI vs AGI vs ASI: what's the difference?: the three terms side by side.
- What is superintelligence, and how close are we?: what forecasters expect, and when.
- What is an intelligence explosion, and is it happening?: the self-improvement loop behind most superintelligence forecasts, with 2026 evidence.
Frequently asked questions
What does SSI stand for in AI?
SSI stands for Safe Superintelligence Inc., the AI research lab Ilya Sutskever co-founded in June 2024 with Daniel Gross and Daniel Levy. Its stated goal is one product: a safe superintelligence.
Is SSI the same as SI?
No. SSI is a private company. SI is the term a 29 September 2026 US executive order tells federal agencies to use instead of AI, and it covers the same technology as the existing legal definition of AI.
Who is the CEO of Safe Superintelligence?
Ilya Sutskever has been CEO since co-founder Daniel Gross left on 29 June 2025. Daniel Levy is President, according to SSI's own updates page.
Has SSI released a model?
No. As of 26 October 2026, SSI has published no model, API, paper or product. An investor claimed a model would ship in August 2026, but nothing appeared on SSI's official channels.
How much is Safe Superintelligence worth?
TechCrunch reports a $32 billion valuation from SSI's 2025 round. No newer valuation has been confirmed. Nvidia invested in July 2026, but neither company disclosed the amount; Bloomberg reported $5 billion.
Can I use SSI's AI today?
No. SSI has no chat app, API or waitlist. It says it will not ship anything until it reaches its one goal, so it has no place in a startup's tool stack for now.
Sources
- Safe Superintelligence Inc. homepage
- SSI updates
- Safe Superintelligence Inc. (Wikipedia)
- Ilya Sutskever's Safe Superintelligence Inc. and NVIDIA Announce Long-Term Strategic Partnership
- Ilya Sutskever's Safe Superintelligence partners with Nvidia to scale its AI research (TechCrunch)
- Ilya Sutskever: TIME100 AI 2026
- SSI could be set to unveil revolutionary language model (Globes)
- Inaugurating the Era of Super Intelligence (White House executive order)
- Salesforce CEO's 11-word post reveals a new pressure on tech giants (TheStreet)
- Mark Zuckerberg overhauled Meta's AI org: Meta Superintelligence Labs (Fortune)