Lasso Security

• Platform security vendors such as SentinelOne, which agreed to acquire Prompt Security, could bundle comparable AI runtime protection into suites enterprises already license and shrink the standalone budget Lasso sells into.
• Model providers could build security controls into their own platforms at little or no cost.
• Lasso’s detection accuracy, latency and LEAP throughput figures are its own benchmarks, and an independent evaluation showing weaker results would undercut its pitch to buyers comparing guardrails.
• Lasso limits its documentation to signed-in users, so evaluators who want to inspect the product before talking to sales could favor rivals that publish their documentation.
• Calcalist reported acquisition offers to Lasso’s cohort from Check Point, Zscaler and F5, and a sale would fold Lasso’s product into an acquirer’s platform.

Lasso Security sells to enterprises and government agencies that want to see and control how employees and AI agents use AI models. Lasso says it discovers agents built on platforms such as Vertex AI, Microsoft Copilot, AWS Bedrock and Salesforce Agentforce, masks sensitive data in prompts before they reach third-party models, and enforces policy inline on agent and application traffic. SecurityWeek described the company at its stealth exit as a Tel Aviv startup building technology against the cyber threats that generative AI and large language models face.

Customers describe the problem in Lasso’s case studies. In the Optibus case study, the customer says its first goal was to see how AI was used across the company. Lasso says the deployment eliminated shadow AI across more than 300 employees and cut AI-related security events by 60%. Lasso’s case study on Telit Cinterion, an IoT company with about 1,000 employees, describes the customer embracing generative AI while maintaining robust security standards.

Independent reporting documents the underlying exposure through research Lasso’s own team published. TechCrunch reported in February 2025 that Lasso found more than 20,000 since-private GitHub repositories whose data stayed reachable through Microsoft Copilot. The Register reported in 2024 that businesses had published code incorporating a software package that generative AI invented, and Lasso researcher Bar Lanyado named Alibaba among them. [s6, s5, s2, s28, s9, s11, s32, s33]

The Lasso platform runs from discovery through runtime enforcement. Lasso says it discovers agents through platform integrations and CI pipelines, maps their models, system prompts, tools and guardrails, assesses posture against NIST frameworks and the OWASP Top 10, red-teams applications, and enforces policy inline at the proxy, API or AI gateway layer.

Intent Deputy is the engine Lasso uses to judge agent behavior at runtime. Lasso says it builds a behavioral baseline for every AI user and agent, analyzes the reasoning behind each action in real time, and can block or quarantine an action before it executes or alert the security team. On its homepage, Lasso cites a 98.6% threat detection accuracy rate. In September 2026 Lasso launched LEAP, a transformer-free guardrail it says returns a verdict in under five milliseconds without specialized hardware, with a self-hosted model called RAPID handling the harder policy calls. SiliconANGLE notes that those benchmarks are Lasso’s own.

Red teaming and open-source code round out the capability set. Lasso says its static attack library adds more than 500 new variants a week, it cites a library of more than 300,000 attacks on its platform page, and it has released an open-source MCP gateway that proxies agent connections to tools. Lasso limits its documentation to signed-in users. [s1, s2, s6, s46, s13, s36, s4, s22, s26, s45]

Lasso competes in a crowded field of AI security startups. Calcalist reported in August 2025 that SentinelOne was acquiring Prompt Security in a deal valued at about $250 million, and that Lasso, Aim Security and Pillar Security had received acquisition offers from Check Point, Zscaler and F5. CB Insights lists Guardrails AI, HiddenLayer, Arthur, Cranium and Wiz among Lasso’s competitors.

Lasso’s stated edges are intent-based detection and the cost of running it. Calcalist reported in 2024 that Schulman named an in-house language model, trained to detect intrusion attempts without slowing software, as Lasso’s main selling point. Lasso now pitches LEAP on cost, arguing that the industry habit of putting a reasoning model in front of every AI request is too expensive at enterprise scale.

Lasso sells against platform vendors as well as startups. Lasso reports that Gartner’s AI usage control overview lists it alongside Microsoft, Cisco, Zscaler, Palo Alto Networks and Check Point. [s31, s30, s38, s29, s13, s24]

Lasso names its customers publicly. On its product pages, Lasso publishes a trusted-by list naming the U.S. Department of Homeland Security, Optibus, Delek US, Compugen, Kaufman Rossin, Artlist, Telit, Kaltura, eToro and Nayax, and its case studies quote Optibus’s Global CISO and Glide Identity’s head of application security. In its September 2026 funding announcement, Lasso named the Department of Homeland Security, Kaufman Rossin, eToro and Fiverr as representative customers and quoted the Kaufman Rossin CISO.

Lasso supplies its own traction figures. Lasso says revenue grew more than 500% in the 12 months before September 2026. SiliconANGLE relayed that claim and noted that Lasso gave no absolute revenue number for that period. Lasso also says LEAP runs in production at global enterprises and the U.S. federal government. In December 2024, Calcalist reported sales of nearly $1 million and clients including U.S. federal agencies, while Schulman told the paper that revenue had passed seven figures.

Lasso also sells through marketplaces and partners. Its AWS Marketplace listing for the LLM gateway publishes a $50,000 annual commitment, Lasso announced Azure Marketplace availability in June 2025, and its partner program recruits resellers, MSPs and MSSPs. Lasso formed a partnership with Swish Data to deliver the platform to U.S. federal civilian agencies, the Department of Defense, and state, local and education organizations. [s5, s9, s10, s13, s36, s29, s18, s19, s21, s20]

Four co-founders lead Lasso, with Elad Schulman as CEO, Ophir Dror as chief product officer, Lior Ziv as CTO and Yuval Abadi as COO. Calcalist reports that Schulman returned to founding after two successful exits, and Lasso’s team page credits Ziv with more than a decade of experience in cybersecurity, generative AI and machine learning.

Lasso’s research team publishes steadily, and independent outlets cover the work. The Register reported security researcher Bar Lanyado’s package-hallucination study in 2024, TechCrunch quoted Dror on the Copilot findings in 2025, and in September 2026 The Register reported Lasso’s finding that AI watermarking changes how agents handle tools and safety refusals. On its research page, Lasso also lists a 2026 disclosure, CVE-2026-77521, on bypassing an agent sandbox in MaxKB.

Lasso’s board brings operating and security experience. On its team page, Lasso lists former Israeli Prime Minister Naftali Bennett, a former CEO of Cyota, which RSA acquired, and Axonius co-founder Dean Sysman, and Calcalist reports that Bennett joined the board shortly after a fundraising round. Lasso’s September 2026 announcement quotes board member Ryan Crum, a ClearSky partner and former CISO of Apollo Global Management. [s7, s29, s33, s32, s34, s35, s8, s15, s13]

Lasso says it is SOC 2 Type II certified for the Security criteria, which it first announced as SOC 2 Type 2 compliance in a February 2024 blog post. Its site footer shows badge images labeled ISO, AICPA, GDPR and PCI.

Lasso supports its federal claims with its own materials. The Department of Homeland Security appears on its trusted-by list. Lasso publishes an anonymized case study of a federal agency that administers lawful immigration, and Lasso says in the Sysman board announcement that enterprises and federal agencies already trust the platform. Lasso tells government agencies on its public-sector page that it offers in-line policy enforcement and audit-ready records, and its posture module maps a customer’s AI environment against NIST frameworks and the OWASP Top 10. [s13, s17, s1, s5, s12, s15, s41, s2]

Add analyzed competitors to compare them side by side with Lasso Security.

Lasso targets enterprises and government agencies whose employees and developers adopted AI faster than their security teams could govern it. Its trusted-by list mixes the U.S. Department of Homeland Security with companies such as Optibus, Telit, eToro and Kaltura, and SiliconANGLE reports, citing Lasso, that the software runs inside customer environments across financial services, insurance, hospitality and automotive.

Lasso segments the product by the AI surface it protects. It sells usage control for employee AI tools, security for agents built on platforms such as Vertex AI and Microsoft Copilot, and protection for customer-facing chatbots and back-end AI applications. Lasso lists an employee product and an LLM gateway separately on AWS Marketplace.

Lasso’s growth plans name federal and regulated industries and two regions. It says it will use its September 2026 funding to deepen federal and regulated-industry work and expand sales in North America and Europe. Lasso offers separate US and EU log-ins to its platform.

Lasso’s capabilities form one lifecycle of discovery, assessment and protection. Lasso says it inventories agents and applications through platform integrations and CI pipelines, maps their models, system prompts, tools and guardrails, red-teams them with multi-turn attacks, and enforces policy inline at the proxy, API or AI gateway layer.

Lasso’s claimed AI advantage is intent-based detection at low cost. Its Intent Deputy engine builds a behavioral baseline for every AI user and agent and flags deviations. Calcalist reported in 2024 that Lasso CEO Elad Schulman called an in-house language model trained to detect intrusion attempts Lasso’s main selling point. LEAP, launched in September 2026, is a transformer-free guardrail that Lasso says decides in under five milliseconds without specialized hardware, while RAPID, a self-hosted model, handles the decisions that need plain-language policy judgment.

Independent outlets have covered Lasso’s research, while its product performance figures remain its own. SiliconANGLE writes that the LEAP benchmarks are Lasso’s own. On its blog, Lasso quotes Latio’s 2026 market report praising its intent-based runtime protection.

Lasso runs a sales-led enterprise motion and opens conversations with a free assessment. On its free assessment page, Lasso offers vulnerability discovery mapped to the OWASP Top 10 for LLMs and agentic applications. In its September 2026 announcement, Lasso says that its offensive team red-teams customers’ AI applications and agents and feeds what it finds back into the product.

Lasso sells through cloud marketplaces and partners as well as directly. It lists products on AWS Marketplace, announced Azure Marketplace availability in June 2025. It recruits resellers, MSPs and MSSPs into a partner program, and partnered with Swish Data to deliver the platform to U.S. federal civilian, defense, and state and local agencies.

Lasso raised money in September 2026 to expand the motion. ClearSky Advisors led a $30 million round, according to SiliconANGLE, and Lasso says it will spend the funding on engineering, federal and regulated-industry work, and go-to-market in North America and Europe.

Calcalist reported in August 2025 that SentinelOne was acquiring Prompt Security in a deal valued at about $250 million, and that Lasso, Aim Security and Pillar Security had received acquisition offers from Check Point, Zscaler and F5.

Lasso prices its LLM gateway on AWS Marketplace at a $50,000 annual commitment. Buyers commit for a one-year base term and pay an additional usage fee when consumption exceeds what the commitment covers.

Lasso lists the employee product and the LLM gateway as separate AWS Marketplace offerings. Lasso does not say how the separate products combine in an enterprise contract, so a buyer pricing the whole platform would need a sales conversation.

Lasso argues for its products on cost. It argues that putting a reasoning model in front of every AI request is too expensive at scale. Lasso pitches LEAP as requiring no specialized hardware and says RAPID costs 100 to 200 times less than calling a cloud LLM API. Those comparisons are Lasso’s own.

Lasso deploys through several paths that match its product split. It says its browser extension deploys across major browsers in minutes. Lasso says its platform deploys through a gateway, API or SDK with a single line of code, and that its platform enforces policy at the proxy, API or AI gateway layer alongside existing infrastructure. In Lasso’s case study, Optibus started with the web extension and plans to add Lasso’s AI gateway for its customer-facing services.

Lasso pitches LEAP as deployable wherever a customer’s data must stay. It says LEAP needs no specialized hardware and can run in a customer’s own cloud, in regulated environments or in air-gapped networks. SiliconANGLE reports, citing Lasso, that the software covers tens of thousands of agents and billions of requests and actions a month.

Lasso limits its documentation to signed-in users. A visitor following its documentation link reaches a sign-in page with separate US and EU instances, so a buyer judging deployment effort cannot read the documentation without an account.

Lasso says it is SOC 2 Type II certified for the Security criteria, after announcing SOC 2 Type 2 compliance in February 2024. Its site footer shows badge images labeled ISO, AICPA, GDPR and PCI.

Lasso supports its federal claims with customers and partners it names. The Department of Homeland Security appears on its trusted-by list, Lasso publishes an anonymized case study of a federal agency that administers lawful immigration. Swish Data delivers the platform to public-sector buyers. Lasso says LEAP runs in production at the U.S. federal government. Lasso tells government agencies on its public-sector page that it offers in-line policy enforcement and audit-ready records.

Lasso builds on the platforms its customers already run. Lasso connects to agent builders such as Vertex AI, Bedrock and Salesforce for discovery. For usage control, it connects with security stacks such as Palo Alto Networks and CrowdStrike. Lasso has also made its integration with Claude’s Compliance API generally available.

Lasso publishes some of its code openly. Its MCP gateway is an open-source proxy for agent connections to tools, released on GitHub under the MIT license, so developers can adopt and inspect that component without buying the platform.

Lasso distributes through marketplaces and partners. It sells on AWS Marketplace and Azure Marketplace, recruits resellers and service providers, and works with Swish Data in the public sector.

The four co-founders cover the core executive roles, with Elad Schulman as CEO, Ophir Dror as chief product officer, Lior Ziv as CTO and Yuval Abadi as COO. Calcalist reports that Schulman launched Lasso after two successful exits, and Lasso’s team page credits Ziv with more than a decade in cybersecurity, generative AI and machine learning.

Lasso’s researchers publish work that independent outlets cover. The Register covered Bar Lanyado’s 2024 package-hallucination research, TechCrunch quoted Dror on the 2025 Copilot repository findings, and The Register reported Lasso’s 2026 research on how AI watermarking changes agent behavior.

The board adds operating and security experience. Calcalist reports that Naftali Bennett, a former Israeli prime minister whom Lasso’s team page lists as a former CEO of Cyota, joined the board shortly after the seed round. Lasso also lists Axonius co-founder Dean Sysman, and its September 2026 funding announcement quotes board member Ryan Crum, a ClearSky partner and former CISO of Apollo Global Management.