Israel already has regulatory sandboxes. Here's what's missing.

Israel has more operational AI regulatory sandboxes than most people — including most Israelis — realize. The narrative that Israel lacks regulatory infrastructure for AI innovation is outdated. What's missing is more specific than that.

Start with what exists. Israel's autonomous vehicle regulatory sandbox is operational and producing real-world results. In Nahariya, Bus Line 5 runs autonomously under a Ministry of Transport pilot, powered by Imagry's mapless driving software. The system operates on an 8 km urban route through the city's busiest areas, with a safety driver present as part of the pilot's progression toward fully driverless operation. In 2022, the Knesset passed legislation enabling commercial driverless vehicle operations — making Israel among the first countries globally to create a comprehensive legislative framework for AV deployment. A December 2025 academic paper in Cambridge Core documented how Israel's regulatory sandbox framework grants specific exemptions while maintaining safety oversight through control centers and mandatory incident reporting.

In healthcare, the Innovation Authority launched a Regulatory Pilot Program in January 2026 for breakthrough AI technologies. The program combines financial support with a dedicated regulatory sandbox, enabling testing in real clinical environments — hospitals, community clinics, virtual settings — with active regulator guidance from the Ministry of Health. The call for proposals, with a deadline of April 14, 2026, specifically targets AI systems with high levels of automation and autonomy in healthcare.

The general framework exists too. In January 2025, the government published legislative guidelines for regulatory sandboxes — a technology-neutral four-phase lifecycle framework applicable to any sector. The architecture for fintech sandboxes is advancing through a multi-regulator committee structure. Israel's National AI Program explicitly identifies "Technological Disruption & Sandboxing" as a core pillar, navigating what it describes as the inherent tension between the swift pace of innovation and the lengthy process of developing regulation.

What doesn't exist is a Physical AI safety sandbox. No controlled environment where a robot OEM can bring a novel safety architecture — an AI-based safety function, a hardware-enforced safety interlock, a redundant safety hardware layer — and test it under regulatory supervision with real-time feedback. No process for a startup building safety silicon for robots to get pre-market regulatory feedback before committing to a full certification pathway.

This matters because of the numbers. Israel has 123+ Physical AI companies, according to the Innovation Authority's own April 2026 report. Most of these companies sell into Europe. The EU Machinery Regulation 2023/1230 applies from 20 January 2027. A Physical AI safety sandbox would let these companies test compliance approaches, validate safety architectures, and build the regulatory track record that certification bodies need — before the deadline, not after.

Israel has the ecosystem. It has the sandbox architecture. It has operational examples proving the model works. It just hasn't pointed it at Physical AI safety. That's the gap.