On Aug. 13, 2026, Israel's National AI Directorate — a unit inside the Prime Minister's Office — published the implementation plan for Government Resolution 4255, "Accelerating Artificial Intelligence in Israel and Establishing Global Leadership," which the cabinet approved June 16, 2026. The plan sets a five-year vision the Directorate calls "Lived AI": rather than compete to build the largest foundation models, the government says Israel's advantage "will come not from building the largest AI models, but from being the first country where AI is fully integrated into daily public life." It commits to a National AI Institute, five regional "acceleration reactors," a national computing base of at least 100,000 advanced AI accelerators (with an initial 5,000-unit buildout), a domestically built quantum computer, and a push toward chip fabrication at 5-nanometer scale or below, according to the Directorate's own Aug. 13 release.
The plan's most consequential line for governance purposes isn't infrastructure, though. It's the creation of a National Ethics Council to advise on algorithmic transparency, fairness and privacy as AI spreads through public services — paired with the conspicuous absence of anything resembling a binding AI statute.
An Advisory Body, Not a Law
That absence isn't an oversight; it continues where Israeli AI policy has stood since a 2021 Bennett-Lapid government resolution first tasked the Ministry of Innovation, Science and Technology with the brief. A peer-reviewed survey in the Cambridge Forum on AI: Law and Governance describes the resulting regime as "non-binding soft law" — policy papers and sectoral guidance rather than an overarching AI act, with only autonomous vehicles carved out for binding rules under a 2022 traffic-ordinance amendment. The Aug. 13 plan doesn't change that architecture. Each of the plan's regional "reactors" is expected to generate what the Directorate calls "deliberate regulatory friction," reporting barriers it hits so a dedicated team can "translate findings into policy updates" — but the Ethics Council itself gets an advisory mandate, not rulemaking power, enforcement authority, or a statutory basis.
The Steelman for Waiting
There's a real case for that restraint. AI harms are still poorly understood and the underlying models are evolving monthly; locking in binding definitions and thresholds now risks rules that make sense today and look counterproductive in two years. A soft-law body that can iterate quickly, publish guidance, and route real-world friction back into policy has genuine advantages over a legislature writing rules it can't easily revise. Proportionate regulation, done right, sometimes means not regulating yet — and a National Ethics Council focused on transparency and fairness is a reasonable first instrument for a government still learning where the actual harms are.
But the Centralization Predates the Plan
What makes the Aug. 13 announcement worth scrutinizing isn't the Ethics Council alone — it's what the plan formalizes. On Sept. 25, 2025, the government voted to strip AI policy from the Innovation Ministry and house a new National AI Directorate inside the Prime Minister's Office instead, reversing the 2021 arrangement. Brig.-Gen. (res.) Erez Askal was named its director soon after, without a competitive tender — a process the Civil Service Commission justified on security and urgency grounds, according to Calcalist's Ctech, which also reported the Directorate's roughly NIS 120 million 2026 budget came from cross-cutting budget cuts rather than a dedicated appropriation, with 20 authorized positions. Eleven months later, that same untendered office is the body writing national AI strategy, setting infrastructure procurement targets, and now standing up the ethics function meant to check it.
A Precedent Worth Worrying About
Israel has run this play before. The National Cyber Directorate has operated inside the Prime Minister's Office largely on executive decisions and emergency regulations for roughly 15 years; a National Cyber Defense Bill establishing its first binding legal framework passed only its first Knesset reading on June 8, 2026 — amid wartime pressure, and only after "more than a decade" of failed attempts, in the words of the directorate's own head, Yossi Karadi, who called it "a historic milestone" and noted Israel was finally joining "advanced nations that implemented similar laws about a decade ago." The Jerusalem Post's analysis of the new AI Directorate flags this history directly: both bodies sit in the PMO, both were created by government resolution rather than statute, and the AI Directorate's founding mandate — covering "regulation and legislation" alongside strategy — echoes the cyber directorate's original charter. If AI oversight follows the same trajectory, the Ethics Council could spend a decade as an advisory voice with no enforcement teeth while the Directorate that appointed it keeps setting policy by fiat.
What Should Actually Change
None of this argues for rushing a binding AI statute — that would just trade one risk for another. It argues for closing the accountability gap around the Directorate itself while the soft-law approach runs its course: a competitive process for future leadership appointments, published findings from the Ethics Council rather than closed-door advice to the PM's Office that created it, and a real timeline toward the cross-cutting AI "framework law" the Ministry of Justice has reportedly been weighing — one narrowly addressing algorithmic liability and automated-decision redress rather than regulating AI development itself. Israel's instinct to avoid premature hard law is defensible. Housing both the strategy-setter and its ethics check inside the same unbudgeted, untendered office is not really a regulatory choice — it's an oversight gap, and Aug. 13 made it permanent instead of closing it.