Safe Superintelligence Inc: Sutskever's $32B Lab (2026)
Safe Superintelligence Inc, founded 2024 by Ilya Sutskever, is a $32B AI lab building toward safe superintelligence with no products released as of 2026.
Safe Superintelligence Inc, known as SSI, is an AI research lab founded in June 2024 by ex-OpenAI chief scientist Ilya Sutskever along with Daniel Gross and Daniel Levy. Based in Palo Alto and Tel Aviv, SSI has raised roughly $3 billion total and reached a $32 billion valuation in April 2025, backed by Google, Nvidia, a16z, Sequoia, and Greenoaks. Its only stated product is safe superintelligence itself, and it has shipped nothing publicly through 2026.
Safe Superintelligence Inc (SSI) was founded in June 2024 by Ilya Sutskever, Daniel Gross, and Daniel Levy. Headquartered in Palo Alto with an office in Tel Aviv, SSI is a $32 billion AI research lab (April 2025 funding round) whose sole mission is building safe superintelligence. Unlike OpenAI, Anthropic, or Google DeepMind, SSI has released no products, papers, or benchmarks as of mid-2026, betting that safety research must come before any commercial deployment.
Founded: 2024 · HQ: Palo Alto, CA, USA · Team: 20-30 · CEO: Ilya Sutskever · Funding: $3B total raised across 3 rounds (Sep 2024 $1B Series A led by NFDG/a16z/Sequoia/DST Global/SV Angel; Apr 2025 $2B led by Greenoaks Capital, with Alphabet and Nvidia as strategic investors) · Valuation: $32B (April 2025 round)
About Safe Superintelligence Inc.
Safe Superintelligence Inc, often shortened to SSI, was founded on June 19, 2024 by Ilya Sutskever, Daniel Gross, and Daniel Levy. Sutskever had left OpenAI a month earlier, after co-leading the company's superalignment research as chief scientist. Gross previously ran AI efforts at Apple after Apple acquired his startup Cue, and had also run Y Combinator's AI program. Levy was a researcher at OpenAI working on optimization and large-scale training. The three incorporated SSI as a private US company with its main office in Palo Alto, California, and a second office in Tel Aviv, Israel, giving the company access to two of the world's deepest AI talent pools. Sutskever announced the company with a single message: SSI is our mission, our name, and our entire product roadmap. Unlike every other frontier lab, SSI has never shipped a product, an API, a chatbot, or a research paper. The company's only stated output is safe superintelligence itself, an AI system that exceeds human capability across essentially all cognitive tasks while remaining reliably aligned with human intent. Sutskever has said the company will pursue this through a mix of scaling known techniques and new research directions he has not detailed publicly, rejecting any intermediate commercial product line as a distraction. As of mid-2026, SSI still has no public roadmap, no benchmark results, and no model card of any kind. The last 18 months brought more headline news than product news. In September 2024 SSI closed a $1 billion seed round at roughly a $5 billion valuation. By March 2025 a new round pushed the valuation to $30 billion, and in April 2025 a further $2 billion came in at a $32 billion valuation led by Greenoaks Capital. That same month brought a major compute deal: Google Cloud agreed to supply SSI with TPU chips, making SSI one of Google's largest external TPU customers, a notable departure from the Nvidia GPU norm among AI labs. In June 2025, Meta made an unsuccessful bid to acquire SSI for around $32 billion; when Sutskever refused, Meta instead hired SSI co-founder and then-CEO Daniel Gross, along with former GitHub CEO Nat Friedman, to lead its new Meta Superintelligence Labs. In July 2025, Sutskever stepped in as CEO and Daniel Levy was promoted to President. SSI's total disclosed funding stands at roughly $3 billion across three rounds between September 2024 and April 2025. The investor list includes Andreessen Horowitz, Sequoia Capital, DST Global, SV Angel, and seed fund NFDG from the first $1 billion round, with Greenoaks Capital leading the April 2025 round alongside returning backers. Alphabet and Nvidia are both reported strategic investors, giving SSI simultaneous access to Google's TPU fleet and Nvidia's GPU supply chain. No revenue has ever been disclosed, because no product or API exists to generate any. There is no business model to describe yet. SSI takes no customers, sells no API access, and runs no subscription products. Its balance sheet is venture capital plus compute partnerships, primarily with Google Cloud. Every dollar raised funds compute, salaries for a small research team, and the two offices in Palo Alto and Tel Aviv. The $32 billion valuation is a bet on the team and the eventual research outcome rather than on any near-term cash flow. SSI has stayed deliberately small. Headcount was reported at around 20 employees at the time of the March 2025 funding round and had grown to roughly 27 by the end of May 2026, a fraction of the headcount at OpenAI, Anthropic, or Google DeepMind. Ilya Sutskever leads as CEO and chief scientist, Daniel Levy serves as President after his promotion from Principal Scientist, and the company recruits almost exclusively senior researchers and engineers. The loss of co-founder Daniel Gross to Meta in mid-2025 was the most significant personnel change since founding, and retaining research talent against poaching by Meta Superintelligence Labs is a constant pressure on the team. SSI's mission statement doubles as its name: building safe superintelligence is, in the company's own words, the most important technical problem of our time, and SSI calls itself the world's first straight-shot SSI lab, with one goal and one product. The company treats safety and capability as a single combined engineering problem, planning to advance capability quickly while keeping safety guarantees ahead at every step. No papers, model cards, benchmark results, or responsible scaling policy have been published as of mid-2026, consistent with the company's stance that nothing will be shared until the core goal is reached. Against OpenAI, SSI has no equivalent to ChatGPT, the API business, or Sora, but also none of the product or consumer-support overhead competing for research budget. Against Anthropic, SSI lacks any equivalent to Claude or the enterprise contracts that fund Anthropic's safety work, but Anthropic must balance commercial deployment with safety in a way SSI does not. Against Google DeepMind, SSI's TPU access gives it comparable infrastructure without Alphabet's bureaucracy, though DeepMind benefits from direct integration into Google's products. The sharpest new rival is Meta Superintelligence Labs, built in mid-2025 partly from SSI's own former CEO and other hires, competing directly for the same small pool of frontier researchers SSI depends on. As an Israeli-American company with a Tel Aviv office, SSI sits inside Israel's deep pool of AI and cybersecurity engineering talent while remaining US-incorporated for fundraising and governance. Because SSI has no deployed product or API anywhere, it currently falls outside the practical scope of the EU AI Act's obligations for general-purpose AI providers, and no compliance certifications, data residency commitments, or privacy policy have been published, since there is no live system for any to apply to. The next 12 months will likely be defined by three open questions: whether SSI raises again at an even higher valuation to fund continued compute spend, whether it can hold its roughly 27-person team against continued poaching from Meta and others, and whether it breaks its silence with any public research output at all. Given the company's explicit refusal to ship anything before reaching its namesake goal, even a small research disclosure would be read as a major signal about how close SSI believes it is.
Mission
Building safe superintelligence (SSI) is the most important technical problem of our time.
Products
- Safe Superintelligence (unreleased) (Unreleased research program; SSI's sole stated product): https://ssi.inc
Links
Frequently Asked Questions
What is Safe Superintelligence Inc and what do they build?
Safe Superintelligence Inc, known as SSI, is an AI research lab founded in June 2024 by Ilya Sutskever, Daniel Gross, and Daniel Levy, with offices in Palo Alto, California and Tel Aviv, Israel. Unlike OpenAI, Anthropic, or Google DeepMind, SSI has not released a chatbot, an API, a developer tool, or any consumer product. The company describes itself as the world's first straight-shot SSI lab, with one goal and one product: a safe superintelligence, meaning an AI system that exceeds human capability across essentially all cognitive tasks while remaining reliably aligned. As of mid-2026 there is nothing for developers, enterprises, or consumers to sign up for or access. SSI's entire roadmap is internal research toward that single end goal, funded by roughly $3 billion in venture capital at a $32 billion valuation. Its market position is unique: it is one of the highest-valued AI companies in the world with literally zero public-facing products.
Who founded Safe Superintelligence and who is the CEO?
SSI was founded on June 19, 2024 by Ilya Sutskever, Daniel Gross, and Daniel Levy. Sutskever was OpenAI's co-founder and chief scientist until he left in May 2024 following the board upheaval involving Sam Altman. Daniel Gross previously led AI efforts at Apple after it acquired his startup Cue, and had also run Y Combinator's AI track. Daniel Levy was an OpenAI researcher focused on optimization and large-scale model training. Gross was SSI's original CEO, but in June 2025 Meta tried to acquire SSI outright, was refused, and instead hired Gross to co-lead its new Meta Superintelligence Labs alongside former GitHub CEO Nat Friedman. In July 2025, Ilya Sutskever became SSI's CEO and Daniel Levy was promoted from Principal Scientist to President, the leadership structure that remains in place through mid-2026.
How much funding has Safe Superintelligence raised?
SSI has raised roughly $3 billion across three disclosed rounds. The first was a $1 billion Series A in September 2024 at around a $5 billion valuation, backed by Andreessen Horowitz, Sequoia Capital, DST Global, SV Angel, and the seed fund NFDG. By March 2025 a new round valued the company at $30 billion, six times its prior valuation, and in April 2025 a further $2 billion came in at a $32 billion valuation, led by Greenoaks Capital with returning backers including a16z, Lightspeed, and DST Global. Alphabet and Nvidia are both reported as strategic investors, an unusual combination that gives SSI access to both Google TPUs and Nvidia GPUs. No revenue or ARR has been disclosed, since SSI has no product or API to generate either. In June 2025, Meta made an unsuccessful roughly $32 billion bid to acquire SSI outright, which Sutskever rejected; there is no IPO timeline.
What products does Safe Superintelligence make?
SSI makes no public products. It has no chatbot, no developer API, no enterprise SaaS offering, and no open-source model releases, in deliberate contrast to every other major AI lab. The company's stated position, in Sutskever's words, is that the company's first product will be the safe superintelligence itself and it will not do anything else up until then. There is therefore no free tier, no subscription, and no per-token pricing to compare against rivals. SSI's only output to date has been research conducted internally, none of which has been published as papers or benchmark results. This makes SSI unusual even among well-funded AI startups, most of which release at least research previews or developer tools while pursuing longer-term goals. Anyone looking for a usable AI tool from SSI will find nothing as of mid-2026.
Where is Safe Superintelligence headquartered and how big is the team?
SSI's main office is in Palo Alto, California, with a second office in Tel Aviv, Israel, chosen specifically to tap Israel's deep pool of AI and cybersecurity engineering talent. The company has stayed deliberately tiny: headcount was reported at around 20 employees at the time of the March 2025 funding round and had grown to roughly 27 people by the end of May 2026. That is a small fraction of the thousands of employees at OpenAI, Anthropic, or Google DeepMind, despite SSI's $32 billion valuation. The team is composed almost entirely of senior researchers and engineers, with essentially no product, sales, or marketing staff, reflecting the company's research-only mandate. The most significant personnel change since founding was the mid-2025 departure of co-founder Daniel Gross to Meta. Retaining its small research team against continued poaching from rivals like Meta Superintelligence Labs is an ongoing pressure on the company.
What is Safe Superintelligence's mission or research focus?
SSI's mission is captured in its own name: building safe superintelligence is, in the company's words, the most important technical problem of our time. The company calls itself the world's first straight-shot SSI lab, with one goal and one product. Its stated approach is to treat safety and capability as a single combined engineering problem rather than competing priorities, advancing capability as quickly as possible while ensuring safety guarantees always stay ahead. SSI has not published any papers, model cards, or benchmark results as of mid-2026, and has not released a responsible scaling policy or equivalent framework, unlike Anthropic's RSP or OpenAI's preparedness framework. This silence is itself part of the company's stated strategy: nothing will be shared publicly until the core safe superintelligence goal is reached. The research direction reportedly combines scaling known techniques with new approaches Sutskever has not detailed.
Is Safe Superintelligence compliant with SOC 2, GDPR, HIPAA?
SSI has no published compliance certifications such as SOC 2, ISO 27001, or HIPAA, and no trust center exists. There is no data retention policy, no enterprise zero-retention option, and no documented GDPR data processing agreement, because SSI has no API, no customer accounts, and no product through which customer data could flow. The company has not published a privacy policy or terms of service covering any external product, since none exists for the public to use. Because SSI has no deployed AI system anywhere in the world, it currently sits outside the practical scope of the EU AI Act's obligations for general-purpose AI providers. If and when SSI ships a product, compliance documentation would need to be built from scratch, similar to how OpenAI and Anthropic built out their trust centers only after launching commercial APIs. For now, none of this applies because there is nothing to certify.
Who are Safe Superintelligence's main competitors?
SSI's closest peers are the other frontier AI labs, though the comparison is asymmetric because SSI has no products. Against OpenAI, SSI has no equivalent to ChatGPT, the developer API, or Sora, but also carries none of the consumer-support and advertising overhead that competes for OpenAI's research budget. Against Anthropic, SSI lacks any equivalent to Claude or the enterprise contracts that help fund Anthropic's safety research, though Anthropic must balance commercial deployment pressure with safety commitments in a way SSI currently does not. Against Google DeepMind, SSI's direct access to Google TPUs gives it comparable infrastructure without Alphabet's internal bureaucracy, while DeepMind benefits from tight integration into Google's consumer products. The newest and sharpest rival is Meta Superintelligence Labs, formed in mid-2025 partly by hiring SSI's own former CEO Daniel Gross, and it now competes directly with SSI for the same small pool of frontier AI researchers.