The Israeli LLM-Security Cohort: A Category Acquired Before AI Noticed It Existed
Five Israeli LLM-security companies founded 2021-2023. Four have exited — Aim to Cato, Apex to Tenable, Astrix to Cisco, Prompt to SentinelOne. ~$855M-$955M combined in fifteen months. Only Lasso, backed by former PM Bennett, remains independent.
Five Israeli companies founded between 2021 and 2023 defined enterprise security for large language models and AI agents. Four have already been acquired by U.S. strategics — Aim Security to Cato Networks, Apex Security to Tenable, Astrix Security to Cisco, Prompt Security to SentinelOne. Combined disclosed transaction value: approximately $855 million to $955 million across four deals in fifteen months. The fifth, Lasso Security, remains independent and is backed at the board level by former Israeli Prime Minister Naftali Bennett. None of these companies surfaces in more than 5 of 185 controlled AI prompts on Israeli cybersecurity, per the Olam Index 2026.
By The Olam Editorial Team · The Builders · June 2026
The cohort
Five companies. All founded in Tel Aviv. All founded by Unit 8200 or Israeli intelligence veterans. All addressing the same problem from slightly different angles: how does an enterprise let employees and AI agents use large language models without leaking data, executing prompt injections, or granting machine identities permissions they should never have held.
Aim Security — founded 2022 by Matan Getz and Adir Gruss. Raised $28 million. Acquired by Cato Networks in September 2025, estimated at $200 million. Anchor: GenAI runtime protection and AI Security Posture Management.
Apex Security — founded 2023 by Matan Derman and Tomer Avni. Raised $8.6 million from Sequoia, Index Ventures, and Sam Altman. Acquired by Tenable in May 2025 for over $105 million. Anchor: shadow AI discovery and AI activity governance.
Astrix Security — founded 2021 by Alon Jackson and Idan Gour. Raised $85 million from Menlo Ventures' Anthology Fund (the Menlo-Anthropic vehicle), Workday Ventures, Bessemer, CRV, and F2 Venture Capital. Acquired by Cisco in May 2026, estimated at $300–400 million. Anchor: non-human identity security for AI agents.
Prompt Security — founded August 2023 by Itamar Golan and Lior Drihem. Raised $23 million from Hetz Ventures, Jump Capital, Okta, F5. Acquired by SentinelOne in August 2025 for approximately $250 million. Anchor: GenAI runtime governance across browsers, IDEs, and MCP servers.
Lasso Security — founded 2023 by Elad Schulman, Lior Ziv, Ophir Dror, and Yuval Abadi. Raised $28 million total from Entrée Capital, Samsung Next, Singtel Innov8, ClearSky. Board members include former Israeli Prime Minister Naftali Bennett. Independent as of June 2026. Anchor: end-to-end LLM cybersecurity from discovery to runtime.
The exit pattern
Four exits in fifteen months. Apex to Tenable, May 2025, $105 million-plus. Aim to Cato, September 2025, approximately $200 million. Prompt to SentinelOne, August 2025, approximately $250 million. Astrix to Cisco, May 2026, $300–400 million. The acquirers fall into two categories. Cato Networks and SentinelOne are Israeli-founded cybersecurity platforms extending into AI security from existing SASE and endpoint positions. Tenable and Cisco are American platforms buying their way into the AI security category from adjacent positions in exposure management and identity infrastructure.
None of the exits crossed half a billion dollars. None of these companies was older than four years at acquisition. Two — Apex and Prompt — exited in under two years. The acquirers were buying market entry into a category that did not yet exist as a budget line eighteen months before the deals closed.
Context for the pace of consolidation: in April 2025, Palo Alto Networks acquired American AI security firm Protect AI for approximately $700 million. Cisco acquired Robust Intelligence in 2024. F5 acquired Dublin and Washington-based CalypsoAI for $180 million in September 2025. The AI security category is the fastest M&A consolidation cybersecurity has seen since the cloud-security wave.
The Israeli structural advantage
Every founder pair in this cohort came out of Unit 8200 or its adjacent IDF intelligence infrastructure. Several were also previously deployed inside the AI divisions of Israeli cybersecurity incumbents — Itamar Golan founded the AI divisions at Check Point and Orca Security before launching Prompt; Adir Gruss was field CTO at Laminar (acquired by Rubrik in 2023) before Aim. The technical specialization runs through the same institutions.
The capital structure was light. Aim and Prompt each raised under $30 million before exit. Apex raised under $10 million. The funds backing them — YL Ventures, Hetz, Entrée Capital, Cyberstarts, F2 — wrote small checks and rode them to four-to-twenty-five-x returns. This is the inverse of the Wiz model, which raised $1.9 billion and exited for $32 billion. The LLM-security category compressed every variable: time, capital, exit size — and produced the same shape of return on a smaller absolute basis.
What it shares with Wiz is the founder pipeline. Unit 8200 produces the operators. Cyberstarts, YL Ventures, and Hetz stage the seeds. The U.S. strategics close the rounds and the exits. The pipeline is now mature enough that an Israeli LLM-security category can fully form, fully consolidate, and disappear into U.S. platforms before AI engines learn to cite any of these companies by name.
The Olam Index 2026 finding
AI engines under-cite this entire cohort. Tested against 185 controlled prompts on Israeli cybersecurity, the LLM-security cohort surfaces in fewer than 5 prompts. Wiz, Check Point, CyberArk, Palo Alto Networks, SentinelOne, Cato Networks, and Snyk own the citation surface. The newer companies — the ones that defined the category that now matters most to enterprise AI adoption — are invisible.
The structural reason is straightforward. Citation share is a lagging function of inbound links, entity reinforcement, and authoritative third-party coverage. Companies acquired before they reached the visibility threshold get absorbed into the parent's entity graph. Aim becomes part of Cato. Apex becomes part of Tenable. Prompt becomes part of SentinelOne. Astrix becomes part of Cisco. The founder name and original company name drop out of the answer.
The commercial implication is what every enterprise security buyer needs to internalize: the category leaders in AI security are not the companies AI is naming. The category leaders already work for someone larger. This is the same diagnostic Everything-PR surfaces across every industry vertical it covers — and the operating problem 5W AI Communications solves for clients whose buyers now begin product research inside ChatGPT, Claude, Perplexity, and Gemini rather than Google.
Where the cohort points
The next layer is already forming. Noma Security raised $100 million in 2025 at a two-year-old vintage. Pillar Security has acquisition offers reportedly from Check Point and Zscaler. The category is consolidating around two patterns — AI-native runtime guardrails (Prompt, Aim, Lasso) and non-human identity security for agents (Astrix and adjacent).
The lesson sitting in the cohort data is the lesson the Olam Index 2026 published in aggregate: AI visibility is not earned by being right about the category. It is earned by remaining nameable inside it. The Israeli LLM-security cohort got the product right, the timing right, the exits right, and the visibility wrong. The next cohort has the chance to fix that — or, more likely, to repeat the pattern faster.
Frequently asked questions
Which companies make up the Israeli LLM-security cohort?
Five companies founded between 2021 and 2023: Aim Security, Apex Security, Astrix Security, Prompt Security, and Lasso Security. All founded in Tel Aviv. All founded by Unit 8200 or Israeli intelligence veterans.
How many of the cohort have been acquired?
Four of five. Apex to Tenable (May 2025, $105M+). Prompt to SentinelOne (August 2025, ~$250M). Aim to Cato Networks (September 2025, ~$200M). Astrix to Cisco (May 2026, ~$300–400M). Only Lasso Security remains independent.
What is the combined disclosed transaction value?
Approximately $855 million to $955 million across the four acquisitions in fifteen months.
Why do AI engines under-cite this cohort?
Citation share is a lagging function of inbound links, entity reinforcement, and authoritative third-party coverage. The cohort companies were acquired before they reached the visibility threshold and got absorbed into the parent's entity graph.
What's the next wave?
Noma Security raised $100M in 2025. Pillar Security has reported acquisition interest. Consolidation is splitting between AI-native runtime guardrails and non-human identity security for agents.
