A Consultation Before a Compliance Regime
On Sunday, July 5, 2026, Israel's National AI Directorate — housed in the Prime Minister's Office — sat down at Industry House with representatives of the Israel High-Tech Association at the Manufacturers Association of Israel for the first formal consultation of its kind. Shahar Bracha, the Directorate's deputy director, said the round is meant to help formulate "the model for implementing the National AI Program" and to run consultations "with the entire Israeli market." The stated purpose was narrow and practical: map industry needs, identify strategic partners, and gather input ahead of establishing a National AI Institute gov.il.
That sequencing is worth noting on its own terms. Israel is not opening its AI strategy with a compliance mandate or a licensing regime — it is opening with a listening exercise. For a government eager to avoid the kind of front-loaded, paperwork-heavy AI Act model the EU has pursued, that is a defensible, even sensible, first move.
The Program Behind the Meeting
The consultation follows the government's June 16, 2026 approval of Prime Minister Benjamin Netanyahu's National AI Program, a strategic resolution spanning infrastructure, R&D, human capital, the labor market, public services, and international partnerships. Its headline commitment is a national target of 100,000 processing units for AI applications, alongside a national quantum computer built substantially on Israeli technology, and a concentrated push on Cyber AI, Physical AI, and countering deepfakes JNS.
The Directorate itself is newer than the program: it was established inside the PM's Office in October 2025, with Brig. Gen. (res.) Erez Askal appointed as its head, reporting directly to Netanyahu rather than through the Ministry of Innovation, Science and Technology, which had previously owned tech policy CTech.
The Case for Centralizing Under the PM
There is a real argument for that structure. AI policy cuts across defense, industry, education, and finance — putting one office with direct prime-ministerial backing in charge can cut through the inter-ministerial turf fights that have slowed digital initiatives in plenty of other democracies. A single accountable directorate, empowered to negotiate with industry on Day One rather than waiting for a statute, can move at the speed AI development actually requires. Given Israel's tight timeline — a national compute buildout is not something that tolerates years of legislative drafting — speed has genuine value.
Where the Model Strains
But concentrating authority in the PMO also concentrates the failure modes. Askal's appointment bypassed the standard competitive tender process; the Civil Service Commission cited "urgency" and "security considerations" to justify the shortcut. The Directorate's initial NIS 120 million budget for 2026, plus 20 new hires, moved control of significant AI spending — including contracts with "high-cost external companies" — out of the ministry structure and into an office with less established oversight machinery. Observers have also flagged the involvement of entrepreneur Dovi Frances, described as close to both Netanyahu and his son, in pushing the Nagel Committee recommendations that underpinned the program before cabinet sign-off CTech.
None of this means the National AI Program is corrupt or that industry consultation is cover for something worse. It means the accountability structure around a fast-moving executive initiative is thinner than the ambition of the initiative itself — and that gap, not the substance of AI policy, is the actual regulatory risk here.
The Compute Number Nobody Has Priced
The 100,000-processing-unit target is the program's most concrete number and its least financially grounded one. Analysts estimate that building out a GPU network at that scale would cost somewhere between $20 billion and $30 billion — but that figure comes from outside experts extrapolating from hardware costs, not from the government's own resolution. The official text approved on June 16 does not specify an appropriated budget for the buildout itself Let's Data Science. A more grounded, incremental figure sits alongside it: roughly 5,000 of the most advanced GPUs made accessible to research and industry each year from 2027 through 2032 — a pathway that, unlike the 100,000-unit headline, at least has a phased structure attached to it.
That gap between the aspirational target and the funded pathway is exactly the kind of thing an industry consultation should surface. If Tuesday's meeting at Industry House produces a realistic phasing plan tied to actual appropriations, it will have done its job. If it produces another round of press-release numbers without a Knesset-approved budget line behind them, the Directorate will have spent its credibility on a consultation instead of a commitment.
The Right Instinct, With a Blind Spot
Israel's choice to consult industry before regulating it, and to build compute capacity before mandating compliance, is the proportionate approach — better than the EU's compliance-first posture, and better than doing nothing. But proportionate regulation should apply to the government's own procurement and appointment processes just as much as to industry: a tender bypassed for "urgency," a budget line moved between ministries, and a headline figure with no appropriation attached are the details that determine whether this program becomes a genuine national capability or a well-publicized set of aspirations.