A Cabinet Green Light, Long on Hardware
On June 16, 2026, Israel's government unanimously approved Prime Minister Benjamin Netanyahu's National AI Program, the most concrete articulation yet of a strategy the Prime Minister's Office has been assembling since it created the National AI Directorate the previous autumn. The headline commitment is a national target of 100,000 processing units for government and national use — sovereign compute that outside analysts estimate could cost $20 billion to $30 billion to build out in full (Let's Data Science). Alongside the hardware target sit two flagship institutions: a National AI Institute meant to link government, academia, industry, and investors through "acceleration hubs," and a national quantum computer program built, where possible, on Israeli technology (Israel National News). The decision sits inside a broader multi-year strategy that OECD's policy tracker has logged as a roughly NIS 1 billion program running through 2027 — June 16 is the first unanimous cabinet-level commitment to this specific infrastructure and institutional footprint (OECD.AI).
The Directorate itself is a modest line item next to those ambitions — a NIS 120 million appropriation for 2026, dropping to NIS 13 million a year from 2027, run by Brig. Gen. (res.) Erez Askal, the country's first AI chief (Jerusalem Post). That gap between a small administrative budget and a tens-of-billions infrastructure ambition is the plan's most honest signal: this is a coordinating body setting direction, not an agency writing the actual checks. The capital has to come from the Finance Ministry, private investors, and foreign partners — the kind of arrangement already visible in a separate deal to have Nebius build and operate part of Israel's national supercomputing capacity.
The People Problem Nobody Skipped
To its credit, the program does not treat compute as the whole story. It commits to workforce training "across education, academia, and labor market sectors," and Israeli universities are preparing a new AI degree for launch this October, positioned to rival computer science as the default technical credential (Jerusalem Post). Estimates of how many workers will eventually need retraining as AI reshapes technical and white-collar roles run into the millions — a figure worth treating skeptically given how speculative workforce-displacement modeling still is, but directionally the right instinct: infrastructure spending without a parallel skills pipeline is how industrial policy stalls.
Analysts are already pushing back on the compute target itself. Tel Aviv University's Nadav Cohen has called the 100,000-processing-unit goal "very ambitious" and cautioned that advanced server clusters become obsolete within two to four years — an argument for narrower bets on physical AI and edge systems, where Israel's defense-technology base gives it a durable edge, rather than chasing frontier-model scale it cannot match against the US or China (Let's Data Science). Energy is a separate constraint: a single hyperscale AI facility can draw power comparable to a mid-sized municipality, and Israel's grid capacity has not obviously caught up to the ambition (Calcalist).
Where the Law Hasn't Caught Up
The program's conspicuous silence is on the legal side. Israel still has no AI-specific statute. What exists is a "Responsible Innovation" policy — sectoral, risk-based rules enforced by existing regulators (banking supervision, the Ministries of Health and Education, the Consumer Protection Authority) rather than one horizontal AI Act (IAPP). The Ministry of Justice is separately weighing a cross-cutting "Framework Law" that would leave sectoral rules in place but supply statutory footing for issues that cut across all of them: liability when an autonomous system causes harm, the legal status of AI-generated content, algorithmic discrimination. That internal discussion has run for months without producing draft legislation.
The gap isn't theoretical. On December 31, 2025, the Tel Aviv District Court ruled in Thaler v. Commissioner of Patents (case 33353-05-23) that an AI system cannot be registered as a patent inventor under Israeli law, because the term "inventor" is "intrinsically linked to human agency." The court did not strain the statute to reach that answer:
"Any fundamental shift in intellectual property policy... is a matter for the legislature rather than judicial intervention."
(Herzog Fox & Neeman) That is a court doing exactly what courts should do when a legislature hasn't acted — resolving the case in front of it without inventing new doctrine. But it previews harder questions ahead: liability allocation, discrimination claims, and the evidentiary weight of AI-generated outputs are less amenable to a patient "wait for the legislature" holding pattern than inventorship turned out to be.
The Case for Moving Slowly — and Its Limits
There is a genuine case for the government's current posture. A comprehensive AI Act risks locking in categories and thresholds that go stale before they're even enforced; Israel's own policy thinking holds that "a law, being a rigid form of regulation, may fail to synchronize with the globally evolving regulation," and that premature codification could burden an export-dependent tech sector with compliance costs that outpace any safety benefit (IAPP). Given how much of the EU AI Act's early implementation has already needed amendment and delayed timelines, that caution is not unreasonable, and this publication has argued the same case against premature horizontal AI statutes elsewhere.
But "not yet" cannot mean "indefinitely," and the National AI Program raises the cost of further delay. A state prepared to spend tens of billions of dollars on sovereign compute and stand up a national institute for AI commercialization needs a predictable liability regime to make the resulting products investable — foreign capital and enterprise customers price legal uncertainty into every deal they sign. A narrowly scoped Framework Law focused on liability allocation and content-provenance rules, layered on top of the existing sectoral regulators rather than replacing them, would give Israel's AI push the legal scaffolding its industrial ambitions already assume exists. The infrastructure decision was unanimous and immediate. The legal foundation underneath it should not be left to drift through another year of internal Ministry deliberation while the compute build-out proceeds regardless.