Former NSO Group CEO Shalev Hulio: Urges Governments to Build Their Own AI Models
Shalev Hulio, former CEO of NSO Group, is promoting 'sovereign AI' through his AI cybersecurity company Dream. He advocates for narrow models running on one to three GPUs to avoid reliance on US cloud providers. Dream has raised $400 million and is valued at $3 billion, with clients in over a dozen countries across Europe, Asia, and the Middle East. However, Singapore’s Coordinating Minister for National Security K. Shanmugam publicly stated that governments do not need to build their own models for every use case; instead, they should evaluate and adapt existing best-in-class models.
Shalev Hulio, former CEO of the Israeli spyware company NSO Group, now runs an AI cybersecurity firm called Dream, which he co-founded in 2022 with former Austrian Chancellor Sebastian Kurz. Dream sells 'sovereign AI' solutions to various governments—using narrow, specialized models that run on one to three GPUs, trained on customer-specific data, without relying on external cloud infrastructure.
Hulio’s core argument is straightforward: AI is critical infrastructure, and governments should not risk the scenario where 'someone wakes up one day and shuts off your model.' In an interview with the UK’s Financial Times, he said, 'Every government should have its own model.' This statement comes from someone actively selling models to governments, making the conflict of interest obvious, but the argument points to a genuine procurement concern—if national security data runs on US companies’ clouds, the supplier’s willingness and availability become external dependencies.
Dream’s market data supports the existence of demand in this sector: the company has raised $400 million in funding and was valued at $3 billion earlier this year, having sold products to more than twelve countries across Europe, Asia, and the Middle East. Dream also gained attention for detecting the 'world’s first end-to-end autonomous AI attack,' reportedly involving Chinese hackers using open-source AI models to discover and chain vulnerabilities to breach Taiwan government systems, including those of the nuclear safety agency.