On August 13, 2026, Israel's National Artificial Intelligence Directorate published the National AI Strategic Plan, the implementing document for Government Resolution 4255, "Accelerating Artificial Intelligence in Israel and Establishing Global Leadership," which the cabinet approved on June 16, 2026. The five-year plan sets a national computing base of at least 100,000 advanced AI accelerators, starting with an initial buildout of 5,000. It also commits to a domestically built quantum computer, a chip fabrication facility at 5nm or below, a dedicated AI energy plan, and a feasibility study on AI computing in low Earth orbit. The plan does not propose an EU-style AI Act.
What the plan actually commits to
The organizing idea is "Lived AI". Israel does not try to out-train the largest foundation-model labs. It aims to be the first country where AI is fully integrated into daily life, from healthcare and education to government services. The plan describes five "acceleration reactors": regional labs, each anchored to an existing centre of excellence and paired with a committed business or government lead, taking projects from idea to national-scale rollout (PR Newswire release).
The hardware goals are the headline. The plan pairs a "state as customer" model for the quantum computer with mechanisms meant to secure priority access to output from a domestic advanced fab, according to Unite.AI's summary of the plan. Unite.AI also reports that the document does not state a total budget. Without a funded schedule, the 100,000 figure is an aspiration and not yet a procurement commitment. That is the plan's most important gap.
The plan's own text is on the Israeli government's site: Israel's National AI Strategic Plan. Unite.AI reports that its public consultation closed on September 6, 2026.
The case for a dedicated AI law
The strongest argument for a horizontal AI statute is predictability. A single law with defined risk tiers gives companies and citizens one rulebook. It lets regulators act before harms accumulate, and it avoids a patchwork in which each ministry improvises. Where AI moves into public services, as Israel intends, the stakes rise, because a flawed benefits or triage system affects people who cannot opt out. The concern is real, and the plan itself acknowledges it: it sets up a National Ethics Council to address algorithmic transparency, fairness and privacy across public-sector adoption, per the launch release.
Why the sectoral model is the better fit
Israel's choice is not new. In December 2023 the Ministry of Innovation, Science and Technology published its AI regulation and ethics policy. According to the IAPP's analysis, it says Israel will "forgo formal artificial intelligence legislation". Enforcement stays with existing regulators, such as the health and education ministries, the Banks Commissioner and the Consumer Protection Authority. A coordinating AI Knowledge Center inside the ministry harmonizes their approaches. The IAPP describes the aim as an approach that is "adaptable, considerate and not excessively daunting".
The 2026 plan builds on that foundation instead of replacing it, and there are three reasons that is sound.
First, sector regulators understand their own risks. A bank supervisor knows what a credit-scoring model can break. A general AI statute has to define "high risk" in the abstract, and definitions written in one year age quickly.
Second, the plan's operating principle is proportionality. Per Unite.AI, each reactor project must document the regulatory barriers it meets and propose fixes. Its long-term "right of way" principle holds that actions AI performs better than humans should not face restrictions absent proven material risk, while staying compatible with GDPR and Western privacy standards. This is a feedback loop that turns real deployments into regulatory evidence. A horizontal act front-loads the rules before that evidence exists.
Third, Israel is a small market that sells into large ones. Its firms already have to meet EU and US requirements to export. A bespoke domestic regime would add cost without adding much protection for Israeli users.
Where the plan is exposed
The risk in a sectoral model is gaps and inconsistency, and a soft-law approach depends on regulators actually being resourced and willing to act. If "Lived AI" pushes automated systems into welfare, health and policing, the Ethics Council's role needs to be more than advisory. Published impact assessments, an appeals route for people affected by public-sector AI decisions, and clear limits on surveillance uses would put the ethics commitment on firmer ground.
The plan's sovereignty framing also deserves scrutiny. The Electronic Frontier Foundation argues that digital sovereignty should mean user autonomy, open tools and resilience, not greater state control over citizens' data and communications. It also warns that encryption backdoors are incompatible with real data sovereignty. Israel's plan is mostly about infrastructure and industrial capacity. Even so, it should hold itself to that standard when national security and AI overlap.
On infrastructure, the fab and the quantum computer are the riskiest items. Advanced fabs cost far more than software programs, and the plan states no budget. Guaranteeing priority access to domestic output could raise questions about supplier neutrality. The orbital-computing item is a feasibility study, which is the right level of commitment for an unproven idea.
The takeaway
Israel has chosen to compete on adoption, compute and regulatory agility instead of on rulebook size. That avoids the trap of regulating a technology before knowing where it breaks. The plan will be judged on three things: whether the accelerators are funded and installed, whether the reactors actually surface and remove regulatory barriers, and whether public-sector AI comes with real transparency and redress. If it delivers on those, a sectoral regime is a credible alternative to the EU model. If it does not, the absence of a horizontal law will be an easy scapegoat.