By Hamza Faisal
The man who built one of the world's most notorious spyware tools now wants governments to trust him with their artificial intelligence. Shalev Hulio, who co-founded NSO Group, the Israeli firm behind the Pegasus spyware, has told the Financial Times that every government should run its own AI models rather than depend on American tech giants.
"Every government should have their own model," Hulio told the FT. "They should not take the risk that somebody will wake up one day and shut their model because AI is a critical infrastructure. You want to control your own AI."
The pitch comes from his new venture, Dream — formally Dream Security — a Tel Aviv-based company selling sovereign AI and cyber-defence tools to states. A $260 million funding round valued Dream at $3 billion, tripling its valuation in four months, and the company has raised more than $400 million in total, with backers including Bain Capital Ventures and Abu Dhabi-based Tau Capital.
From Pegasus to sovereign AI
Hulio led NSO Group for a decade before leaving in 2022. Pegasus, the company's flagship spyware, was used to hack the phones of journalists, activists and officials around the world, and NSO was later blacklisted by the US government and lost a lawsuit brought by WhatsApp over Pegasus-related hacks. Hulio's argument now is that the same governments he once sold surveillance tools to should buy independence instead: models trained on a state's own data, running on its own infrastructure, for tasks such as intelligence analysis and procurement.
Dream has grown to about 350 employees in three years, with offices in Vienna and Abu Dhabi alongside its Tel Aviv base. The company says it has sold cyber-security tools to more than a dozen countries, and that its sovereign AI product is already deployed in two.
What Dream actually sells
Dream's first product, Sphere, provides AI-enabled cyber defence for governments and critical infrastructure. Its second, Hero, is built to find and patch software vulnerabilities autonomously. In August, Dream's researchers said they had found evidence of what Hulio describes as the first end-to-end autonomous attack on a government target — the kind of machine-led threat the company says its tools are built to stop.
The timing has helped the pitch. Israeli publication Globes tied Dream's funding round to Washington's decision to block non-US citizens from accessing some leading American AI models, a restriction that sharpened fears in foreign capitals about depending on US providers for critical technology.
Why the pitch lands — and why it unnerves critics
The sovereignty argument is gaining ground well beyond Dream. Governments worried about data localisation, export controls and shifting alliances increasingly want AI they can host themselves rather than rent from a handful of US cloud giants. Dream's answer to the trust question is structural: customers train and run the models on their own infrastructure, so the company itself does not see their data.
Critics, though, note the irony at the centre of the pitch: a former spyware chief asking states to trust his company with their most sensitive systems. For Hulio, that history is the credential — he argues that his past dealings with governments are exactly why they trust Dream now. Whether that trust survives contact with the next scandal is the question his $3 billion valuation now depends on.
By Hamza Faisal
