Three of the 50 entries in the Israeli Defence-AI Deployment Index are classified as military cyber. They sit at L2 · Contract, L1 · Trial and L0 · Claimed respectively, and none reaches L3 · Operational. Among the Israeli military cyber companies the index covers, the strongest published evidence is a contract announcement with an unnamed government customer.
That is an unusual result for a country with the largest cyber-security export sector per capita in the world. The explanation lies in what this index counts and what cyber customers publish.
What this category covers
Israel has hundreds of cyber-security companies. Almost all of them sell to enterprises, and enterprise security is not defence. The index admits a company only where its core product is AI or autonomy applied to defence, which in this category means software sold to militaries, intelligence services or national security agencies as its primary market rather than as an adjacent one. That test removes the commercial cyber sector almost entirely, and it also removes the primes, for the reasons set out in why the Israeli primes are excluded.
Three companies survived that filter at version 0.2. The grading rules and the full index sit alongside this page.
L2 · Contract — Cyabra
Cyabra (founded 2018, Tel Aviv) sells a platform for detecting coordinated online manipulation, bot networks and state-sponsored influence activity. On 3 August 2026 the company announced a multi-year, six-figure contract with an Asia-Pacific intelligence agency, under which the agency would use the platform to detect, monitor and mitigate online disinformation across the digital domain. On 24 August 2026 it announced agreements with two European national security agencies representing more than $500,000 in combined annual revenue.
Both announcements are company-issued and distributed through a newswire. Neither names the buying agency. The index grades Cyabra L2 because publicly announced procurement exists and nothing published describes the platform in operational use.
L1 · Trial — Toka
Toka (founded 2018, Tel Aviv) builds tools that access internet-of-things devices, including networked cameras, for government customers. TechCrunch reported in December 2024 that the Andreessen Horowitz-backed company sells exclusively to state clients and was seeking to expand into US federal agencies, and described the technology as locating cameras within a perimeter and accessing their feeds.
The index grades Toka L1 because the public record establishes what the company sells and to whom it wants to sell, rather than a publicly announced procurement award. This is a category where the absence of a contract announcement should not be read as the absence of a contract; agencies buying access tooling do not issue press releases about it.
L0 · Claimed — Dream Security
Dream Security (founded 2023, Tel Aviv, with offices in Vienna and Abu Dhabi) is the largest company in this category by valuation. The Times of Israel reported that the company reached a $3 billion valuation and aims to help governments own and control their AI, and Globes reported a $260 million raise at that valuation. SecurityWeek covered the same round.
An L0 grade here means only that no independent public source corroborates a specific defence application of the product. It does not mean the company is small, unfunded or overstating itself; the published record establishes scale and ambition clearly. What it does not establish is a named military or defence customer, which is the sole thing the grade measures.
What the distribution shows
Cyber is the clearest illustration in the whole index of the difference between a large sector and a documented one. Three entries, no L3, and the single strongest piece of evidence in the category is a vendor press release describing an unnamed agency in an unnamed country.
Two structural forces produce that. The first is customer secrecy. National security agencies do not confirm what intelligence or offensive tooling they procure, and vendors that disclose the buyer lose the buyer. The result is a category where the customer field is permanently blank, a problem the index treats separately in the undisclosed-customer problem.
The second is definitional. Much Israeli cyber capability sits inside government units, inside the primes, or inside firms whose defence work is a minority line. The index's inclusion test is narrow on purpose, and a narrow test applied to a broad sector returns a short list.
The comparison with Israeli electronic warfare and SIGINT AI is instructive: same secrecy, same small count, same absence of an L3. Both categories reward a reader who treats a low grade as a statement about publishing behaviour.
What would change it
An L3 in this category would require a named outlet reporting that an identified military or agency used one of these platforms operationally. Nothing in the current publishing norms of the sector suggests that is likely, which is itself a finding worth stating plainly.
More achievable is movement within the lower tiers. A government publishing a procurement notice — an EU tender award, a US federal contract listing, a parliamentary answer — would move Toka or Dream Security to L2 immediately. Public procurement databases are the most likely route by which any cyber entry in this index changes grade, and they are checked at each revision.
For Dream Security specifically, a single report from an identifiable publisher naming a defence or military customer would move the entry. The index revises on evidence, not on scale, as set out in the deployment evidence gap.
Limits of this reading
Three entries cannot describe Israeli military cyber. They describe what has been published about three companies that met a narrow inclusion test.
The index holds no technical information about any of these products and makes no claim about whether they work. It also cannot see classified contracting, which in this category is likely to be the majority of the activity. Readers should treat the cyber row of the index as the row where the gap between the public record and the underlying reality is widest, alongside the other thin categories the index reports.
Frequently asked questions
Which Israeli military cyber companies have publicly announced government contracts?
Cyabra is the only entry in this category with publicly announced agreements: a multi-year contract with an Asia-Pacific intelligence agency announced in August 2026, and agreements with two European national security agencies announced later that month. The buying agencies are not named.
Why is a $3 billion cyber company graded L0?
The grade records evidence of defence deployment, not company size or funding. Dream Security's valuation and raise are well documented, but no independent public source names a military or defence customer for its products, which is what an L0 grade describes.
Are Israel's large cyber-security firms in this index?
No. The index admits companies whose core product is AI or autonomy applied to defence. Enterprise security vendors, and primes for whom cyber is one line among many, both fall outside that test regardless of size or Israeli defence heritage.
Sources
- idf.ai, The Israeli Defense-AI Deployment Index v1.0, 24 August 2026
- GlobeNewswire (company announcement), Cyabra secures multi-year, six-figure contract with APAC intelligence agency to deliver AI-powered narrative intelligence, 3 August 2026
- The Manila Times / GlobeNewswire (company announcement), Cyabra secures contracts with two European national security government agencies, 24 August 2026
- TechCrunch, a16z-backed Toka wants to help US agencies hack into security cameras and other IoT devices, 6 December 2024
- The Times of Israel, Valued at $3 billion, Israeli cyber unicorn aims to help governments own and control their AI
- Globes, Dream Security raises $260m at $3b valuation
- SecurityWeek, Dream raises $260 million at $3 billion valuation
Independent publication of idf.ai. Not affiliated with, endorsed by, or connected to the Israel Defense Forces, the Israeli Ministry of Defense, or any government body. Compiled entirely from publicly published sources. No classified, restricted or non-public information. Listed companies may dispute any entry: send the published source that contradicts it and the entry will be amended or removed.