Lasso Security

• SentinelOne's purchase of Prompt Security armed a platform incumbent with a comparable product. Further acquisitions of Aim Security or Pillar Security would put equivalent AI runtime protection inside platforms enterprises already license, compressing the standalone budget line Lasso Security sells against.
• Model providers could bundle security controls into their own offerings at little or no cost, a dual role as competitor and potential acquirer that Schulman himself acknowledges, eroding the paid defense layer between enterprises and the labs.
• The sub-50ms latency and detection accuracy figures behind Intent Deputy and AI detection are vendor-stated, and an independent review repeats them as vendor benchmarks. An external evaluation showing materially worse results would undercut the intent-security positioning.
• Traction claims are own-voice. Reported revenue, federal-agency adoption, and partner growth all originate with the company, and without an independently reported customer or revenue milestone, buyers and investors weigh vendor assertion alone.
• Reported totals raised diverge across data providers, so the capital base behind an enterprise and federal sales motion is unclear, and rivals in the same cohort have disclosed larger rounds.
• Reported buyout interest means the founders may sell to a platform vendor, which would fold the product into an acquirer's stack rather than an independent platform.

Lasso Security sells to enterprises whose employees and applications adopted generative AI faster than security teams could see or control it. The buyer in the company's public materials is the CISO, who faces shadow AI use, sensitive data leaving through prompts, and autonomous agents acting on production systems. SecurityWeek described the company's mission at its 2023 stealth exit as protecting every LLM touchpoint against data exposure and compliance risk.

Named customers describe the pain concretely. Optibus, a transit software company with more than 300 employees, says it started from zero visibility into AI usage, and Telit Cinterion describes employees adopting ChatGPT, Gemini, and Microsoft Copilot ahead of any security control. Both case studies live on the vendor's site, so they carry the vendor's frame, though the named customers speak in their own words.

Independent reporting corroborates that the underlying threat is real and large. Lasso's own research, covered by TechCrunch and CTech, found more than 20,000 once-private GitHub repositories still reachable through Microsoft Copilot, and The Register documented how AI assistants invent installable package names attackers can register. Calcalist separately reports that GenAI security startups, Lasso among them, are fielding buyout offers, a sign incumbents see the demand as durable. [s16, s9, s11, s19, s28, s25, s18]

The Lasso AI Security Platform runs from discovery to runtime enforcement. Its discovery module inventories agents and applications and builds an AI bill of materials, an automated red-teaming module maps risks across MITRE and OWASP attack vectors, and the detection and response module enforces policy inline to stop a harmful agent action before it executes.

Intent Deputy is the platform's newest layer and its central technical claim. The vendor states sub-50-millisecond analysis on the Intent Security page, which cites a 99.83% threat detection accuracy rate, while the AI detection page cites 98.6% accuracy for intent-based zero-day protection. An independent AppSec Santa review repeats these as vendor benchmarks, and no third party has tested either figure in any fetched source.

Lasso supplements its product pages with original research that gives the capability claims outside reference points. Its researchers showed Microsoft Copilot could retrieve once-private GitHub repositories, work TechCrunch covered in February 2025, and earlier showed AI assistants recommend non-existent software packages, covered by The Register. The research page published new work on LLM fingerprinting and agentic data exfiltration into 2026. The company also packages the platform for AWS Marketplace with one-line-of-code integration and secures traffic across existing AI gateways such as Kong, Portkey, LiteLLM, and Envoy. No public product documentation portal was found. [s2, s4, s5, s3, s26, s19, s25, s8, s13]

Lasso Security competes inside a dense cohort of GenAI security startups that platform vendors have begun buying. SentinelOne acquired the closest Israeli peer, Prompt Security, in a deal Calcalist valued at about 250 million dollars in August 2025, and the same outlet reported that Lasso, Aim Security, and Pillar Security then began fielding buyout offers from global vendors. Each such acquisition removes an independent rival and arms an incumbent with a comparable product.

The company's stated edges are its own detection models and the intent framing. Elad Schulman told Calcalist that Lasso trains an in-house language model to detect intrusion attempts without degrading application performance, and the company markets Intent Deputy under an Intent Security banner it coined for agent defense. Breadth is the other claim, since one platform covers employee usage, applications, agents, and code.

Schulman himself names the structural threat. He told Calcalist that the model providers behind popular AI tools are developing their own security features and act as both competitors offering nearly free solutions and potential acquirers. The same logic now applies to the security platforms, since any of them can buy a Lasso peer instead of building the capability. [s22, s18, s17, s3, s2]

Named customers appear with attributed metrics, all on the vendor's own pages. Optibus reports eliminating shadow AI and a 60% drop in AI-related security events across more than 300 employees, Glide Identity says an agent secured by Lasso cut a three-week penetration test to two days, and the Telit Cinterion case study credits Lasso with enabling GenAI adoption at a roughly 1,000-person IoT company. No independently reported customer win appears in fetched press.

Revenue and pipeline claims are the founder's own. Schulman told Calcalist in December 2024 that the company was generating over seven figures in revenue with major deals lined up, while the same article reported sales nearing one million dollars, and the February 2026 board announcement says enterprises and federal agencies already trust the platform. The buyout interest Calcalist reported implies acquirers see commercial substance, but no third party has published Lasso's customer count or revenue.

Channel motion is visible and early. Lasso listed products on AWS Marketplace with contract-based purchasing, runs a partner program for resellers, MSPs, and MSSPs, and claims 6x partner network growth. The investor base has widened beyond the original seed backers Entrée Capital and Samsung Next to include SingTel Innov8, per CB Insights, though reported totals raised vary across providers. [s9, s10, s11, s17, s15, s14, s12, s27, s29]

Four co-founders run the company across CEO, CTO, CPO, and COO seats. Calcalist describes Elad Schulman as returning to founding after two successful exits, Lior Ziv brings over a decade of experience in cybersecurity, GenAI, and machine learning per the team page, Ophir Dror leads product, and Yuval Abadi runs operations. The team page presents all four alongside the 2023 founding.

Board recruitment is the standout governance signal for a company this young. Former Israeli Prime Minister Naftali Bennett, who founded the security company Cyota and sold it in 2005, joined the board after the seed round, and Axonius co-founder Dean Sysman, who ran that company as CEO for its first nine years, joined in February 2026. René Bonvanie, formerly Palo Alto Networks' CMO, joined the advisory board at the seed announcement. No new C-suite appointment appears in the company newsroom.

A named research bench gives the company public technical credibility. The research page lists bylined authors, the team's Copilot findings drew TechCrunch coverage that quoted CPO Ophir Dror, and its AI package hallucination work drew The Register. The researchers kept publishing through 2026, most recently on LLM fingerprinting and agentic data exfiltration. [s6, s17, s15, s19, s25, s8, s7]

Lasso published a SOC 2 Type 2 announcement on its blog in February 2024, where co-founder Yuval Abadi states the company is officially SOC 2 Type 2 compliant. The post gives a generic SOC 2 description rather than naming the auditor or the trust criteria the report covers, and the report itself is not posted, so a buyer still requests it in procurement. Homepage footer seals display ISO 27001, AICPA SOC 2, GDPR, and PCI DSS as images, with no report linked from the footer, and no FedRAMP authorization appears in the public record.

The company sells compliance outcomes beyond what its own SOC 2 covers. The board announcement positions the platform as helping enterprises demonstrate compliance with evolving AI governance requirements, and the Telit Cinterion case study leads with maintaining compliance during GenAI adoption. Mapping customers to frameworks the company has not certified against itself is a common asymmetry at this stage, and the gap is worth tracking.

Research conduct is the trust signal the company controls directly. The team disclosed the Copilot findings to Microsoft in November 2024 before TechCrunch published in February 2025, responsible-disclosure behavior a buyer can verify in independent press. [s21, s24, s15, s11, s19]

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

Lasso Security targets enterprises whose AI adoption has outrun their controls, and the visible customer set crosses industries rather than concentrating in one. Telit Cinterion is a roughly 1,000-person IoT company, Optibus is a transit software vendor with more than 300 employees, Glide Identity is an identity startup, and the February 2026 board announcement adds federal agencies in the company's own voice. The unifying trait is depth of AI adoption rather than sector.

The newest public product and research activity is current, which matters in a market the frontier labs keep reshaping. Intent Deputy launched in early 2026, and the research page published new work on LLM fingerprinting and agentic data exfiltration into 2026. The pitch itself moved with the model generation, from LLM data leakage at the 2023 stealth exit to agent intent security today, the repositioning a vendor must keep showing to stay fit with a moving market.

Persona targeting splits the economic buyer from the deployment surfaces. The CISO owns the purchase in the customer stories that name a buyer, while the packaging, an employee extension, a developer scanner, and an application gateway, maps onto the distinct teams who touch AI. Geographic strategy is not disclosed, with named customers visible in Israel and the United States and federal interest claimed in the United States.

The claimed capabilities cohere into one lifecycle story everywhere they appear. Lasso discovers agents and applications and builds an AI bill of materials, assesses them with automated red teaming mapped to MITRE and OWASP attack vectors, and protects them at runtime with inline policy enforcement and intent analysis. The marketplace listings, the platform pages, and the customer case studies describe the same mechanism set without contradiction.

The claimed AI advantage is detection technology the company built itself. Schulman told Calcalist that Lasso trains an in-house language model to detect intrusion attempts without compromising software performance, and Intent Deputy extends the claim with behavioral baselines learned per agent. The sub-50-millisecond latency and 99.83% accuracy figures attached to that engine remain vendor-stated, and an independent AppSec Santa review repeats the cost and accuracy numbers explicitly as vendor benchmarks rather than independent measurement.

Research is the capability evidence the product pages cannot supply, and it now carries independent corroboration. The team showed Microsoft Copilot could retrieve once-private GitHub repositories, work TechCrunch covered in February 2025, and earlier documented how AI assistants recommend non-existent software packages, covered by The Register in 2024. Publication continued into 2026 on LLM fingerprinting and agentic data exfiltration. Product-level verification is still the gap. The AWS Marketplace listing points to a vendor knowledge base, and the AppSec Santa review is descriptive rather than a measured test, so no public API documentation, sandbox, or independent benchmark lets a buyer check the platform's depth.

Lasso runs an enterprise sales-assisted motion with the founders out front. The site funnels buyers to demo bookings, the AWS listings route larger buyers to private offers, and no self-service tier or free trial exists. Elad Schulman fronts the press, gave Calcalist the company's revenue and strategy story, and appears across the webinar program, founder-led selling that fits a company at an early revenue stage.

Lasso invested in partner and marketplace channels early in its development. The partner page recruits resellers, MSPs, and MSSPs and claims 6x partner network growth, and the AWS Marketplace presence dates to 2024. The investor base has also widened beyond the original seed backers Entrée Capital and Samsung Next to include SingTel Innov8, and the reported totals disagree, with the CB Insights profile listing $14.5 million raised against the $28 million the AppSec Santa review reports. A rival could assemble similar assets, but they show the company building paths to buyers beyond its own sales calls.

Lasso has claimed federal customers since at least December 2024, when Schulman told Calcalist that U.S. federal agencies are strategic customers, yet the federal motion stays the least evidenced part of the strategy. The Sysman announcement says federal agencies already trust the platform and presents his experience with federal customers as a reason for the board seat. No contract vehicle, FedRAMP status, or named agency appears in fetched sources.

Lasso publishes no prices on its own site, which signals negotiated enterprise deals. The AWS Marketplace listing publishes a $50,000 twelve-month annual commitment for the LLM gateway, points larger buyers to private offers through a sales conversation, and meters additional usage beyond the contract at a published per-unit fee whose unit the listing does not define.

The packaging carries the clearer pricing message. Selling an employee extension, an LLM gateway, and a developer scanner as separate marketplace products lets the company price each AI surface on its own and expand accounts surface by surface. Whether a platform-wide bundle exists, and how negotiated enterprise pricing departs from the published baseline, stays behind the private-offer wall.

Premium positioning is implied but not argued publicly. No fetched page compares Lasso's price to a competitor or to the incumbent platforms now adding AI security features, so the company has not yet had to defend a price premium in public, a question that sharpens as larger security vendors acquire comparable capability, as SentinelOne did in buying Prompt Security.

Deployment models map onto the product split. The employee product installs as a browser extension, the gateway product positions between LLMs and data consumers with integration the vendor claims takes one line of code, and discovery connects through platform integrations and CI. A company post describes securing traffic across existing AI gateways such as Kong, Portkey, LiteLLM, and Envoy rather than requiring its own.

Operational claims center on latency, the make-or-break property for inline AI security. The marketplace materials claim high interaction volumes without latency, and the intent page claims analysis in under 50 milliseconds. Both figures are the vendor's own, and no benchmark, SLA, status page, or uptime commitment appears in fetched sources.

Buyer-facing operations collateral is missing. The public record shows no documentation portal, deployment guide, or API reference beyond a knowledge-base entry on the marketplace listing, so evaluating deployment effort requires a sales engagement. For a product whose pitch is frictionless AI adoption, the absence of self-service evaluation material is a notable gap.

Lasso holds a SOC 2 Type 2 attestation, which it announced on its blog in February 2024 with co-founder Yuval Abadi stating the company is officially SOC 2 Type 2 compliant. The homepage footer also carries image-only ISO 27001, AICPA SOC 2, GDPR, and PCI DSS seals, so the company displays a broader posture than the SOC 2 blog post alone implies. The seals are inline graphics with no linked report, and the blog post describes SOC 2 generically rather than naming the trust criteria the audit covers, so the access the platform requests still draws procurement requests. The platform sits inline on AI traffic and reads prompts, code, and agent actions, and no FedRAMP authorization appears in the public record.

Compliance beyond SOC 2 appears as a customer outcome rather than a vendor posture. The company markets help with AI governance requirements, and the Telit Cinterion case study frames the deployment around maintaining compliance during GenAI adoption, but the frameworks the platform maps customers to go past what Lasso has certified against itself. The federal-agency claim, if accurate, implies procurement scrutiny the public record does not yet fully show.

Research conduct is the trust signal the company controls. The team disclosed the Copilot findings to Microsoft in November 2024 before TechCrunch published in February 2025, responsible-disclosure behavior a buyer can verify in independent press.

Lasso positions as a platform, and the module structure supports the claim. Discovery, red teaming, runtime detection and response, and usage control share one inventory and one enforcement plane, so each module deepens the same deployment rather than starting a new one. The Intent Security framing tries to give that platform a category name of its own.

The ecosystem motion is integrate-and-distribute rather than replace. The company describes securing traffic that flows through existing AI gateways such as Kong, Portkey, LiteLLM, and Envoy, sells through AWS Marketplace, and recruits resellers and MSSPs into a partner program with claimed 6x network growth. That posture lowers adoption friction and turns adjacent vendors into channels rather than rivals.

Third-party ecosystem evidence stops there. No public API documentation, developer program, or example of another vendor building on Lasso appears in fetched sources, so the platform ambition currently runs on architecture and packaging rather than network effects.

The founding team covers the four seats an enterprise security company needs. Calcalist describes Elad Schulman as returning after two successful exits to run the company, Lior Ziv leads technology with a decade in cybersecurity and machine learning, Ophir Dror owns product, and Yuval Abadi runs operations and finance. The team page presents the company as founded by these four in 2023.

The named research bench is unusual for the company's size, and its output now carries sustained independent coverage. The research page lists bylined authors, the 2024 AI package hallucination work drew The Register, and the 2025 Copilot findings put the team in front of Microsoft and the trade press through TechCrunch. Publication continued into 2026, so the bench still produces.

Board members and advisors compensate for the missing executive layer. Naftali Bennett brought Cyota founder experience after the seed round, Axonius co-founder Dean Sysman brought category-creation and federal go-to-market experience in February 2026, and René Bonvanie brought Palo Alto Networks marketing pedigree at the seed announcement. No new CRO, CMO, or VP-level hire appears in the company newsroom.