Cyata

• Identity platform vendors and security suites such as Microsoft, Okta, and CrowdStrike could fold AI agent discovery and least-privilege enforcement into products enterprises already buy, undercutting a standalone agentic identity purchase.
• Cyata operates inside Check Point now, so its roadmap and independence depend on Check Point's priorities, and the standalone product could be deprioritized or folded into the wider platform over time.
• Only one customer is named publicly, so buyers who require several current named references had one case study, a logo wall, and an acquirer's conviction to rely on, a thin signal for a company in its first funding round.
• Cloud and model providers could ship native identity controls for the agents built on their own platforms, removing the third-party budget line an independent agentic identity product depends on.
• Buyers choose Cyata largely for the founders' research output, so a slowdown in that disclosure stream or a key departure after the acquisition could erode the main reason a buyer picked it over a bundled competitor.

Cyata treats the AI agents an enterprise runs as a new identity class that traditional access controls were never built to handle. The homepage frames the problem as agents querying data, calling APIs, and acting autonomously across cloud and SaaS, which security and IT teams cannot fully see or govern. The buyer is the security team standing up agents inside core business processes.

Independent reporting corroborates the gap beyond vendor marketing. Calcalist described a new class of digital worker that executes code, queries sensitive databases, and triggers automated workflows while operating outside the identity frameworks applied to human employees or service accounts. That framing establishes the oversight problem as real rather than vendor-invented.

Cyata positions autonomous action as the consequence that matters. The company argues that agents act at scale yet no one watches what they do, and that discovering, recording, and controlling each agent is what closes the gap before an unmonitored agent causes harm. [s1, s4]

The Cyata product is positioned as a control plane for agentic identity built around three jobs. The homepage describes discovering every AI agent in the organization, recording and explaining what each agent does, and controlling agent access through least-privilege policy, human-in-the-loop approvals, and real-time guardrails. It is meant to integrate with existing cloud, SaaS, and identity infrastructure rather than replace it.

The control layer is where the product claims to act. Cyata describes enforcing safer operational defaults such as least privilege and isolation boundaries, and gating sensitive capabilities like secret access and high-privilege tool execution in untrusted contexts. These are the controls a security buyer evaluating an agent-governance product would test against its own agents.

The research output demonstrates the same capability the product sells. Cyata researchers disclosed the Vault Fault flaws in CyberArk and HashiCorp secret managers, LangGrinch in LangChain Core, and three flaws in Anthropic's official Git MCP server, work now recorded in independent vulnerability databases. This original adversarial work on the identity and agent infrastructure the platform governs validates that the team finds the weaknesses the product is built to catch. [s1, s3, s14, s15]

Cyata competes in agentic identity against both independents and the identity platforms moving toward the same work. Token Security, Clutch Security, Astrix Security, and Oasis Security cover overlapping discovery, governance, and least-privilege enforcement for AI agents and non-human identities, while broader identity and security platforms are building adjacent controls. The category Cyata sells into is one larger vendors are validating by building toward it.

Cyata's visible differentiator is the offensive depth of its team, shown through research. The Vault Fault disclosures in CyberArk and HashiCorp secret managers, LangGrinch in LangChain Core, and the flaws in Anthropic's official Git MCP server give the company a public profile larger than its first funding round would otherwise support, and that research brand is harder for a bundled competitor to reproduce quickly.

The structural question is who owns the buyer, and the Check Point acquisition answered it for Cyata. A platform vendor bought the capability rather than letting it stand alone, which is the same pressure every independent in this group faces from identity platforms and cloud providers that already sit inside the enterprise stack. [s1, s6, s14]

Research is Cyata's clearest go-to-market engine, and it points outward rather than at named customers. The Vault Fault disclosures in CyberArk and HashiCorp secret managers and LangGrinch in LangChain Core drew attention to the founders as researchers worth following and opened enterprise conversations. This is demand generation through research rather than evidence of paid deployments.

Named commercial proof is still thin in the public record. The homepage carries one customer reference, an eCapital case study with a video testimonial from its VP of Technology and Compliance, alongside a logo wall whose brands, identifiable as Mercury, Nayax, Trigo, and Seemplicity from their image files, appear without any text naming them as customers. TLV Partners led the seed, which is investor conviction rather than buyer proof.

The acquisition itself is the strongest external signal. Check Point agreed to buy Cyata about seven months after it left stealth, which a larger vendor does only when it sees value, though it leans on an acquirer's judgment more than on a deep public roster of named references. [s8, s14, s9, s5]

The founders pair offensive-security craft with prior operating experience at established security companies. The team are alumni of Unit 8200, Cellebrite, and Check Point, and CEO and co-founder Shahar Tal was a senior executive at Cellebrite and Check Point before starting the company. Israel's Corporations Authority registers the business as Cyata Security Ltd, an active private company incorporated in 2024, and co-founders Dror and Baruch round out a small founding team.

The research record is the team's strongest public signal, and independent vulnerability databases now carry it. Cyata's Vault Fault project found 14 flaws in CyberArk and HashiCorp secret managers, including a HashiCorp Vault code-execution flaw that had been exploitable for over nine years, tracked as CVE-2025-6000. The team also disclosed LangGrinch in LangChain Core, tracked as CVE-2025-68664, and three flaws in Anthropic's official Git MCP server, a pattern of disclosures in the company's own product domain recorded in NVD and cve.org rather than a single covered event.

The acquisition adds outside recognition. Check Point agreeing to buy the company months after it left stealth is the kind of third-party validation that separates a research claim from a research record, and it ties the team's standing to a concrete commercial outcome. [s4, s13, s14, s12, s15]

Cyata's trust posture centers on the data its product handles when it inspects agent activity. The platform is positioned to observe agent actions in real time and enforce policy, which raises the question a buyer asks when a security product sees proprietary AI traffic, and the company answers it by emphasizing least-privilege and guardrail controls rather than broad data collection.

Cyata claims a SOC 2 Type II attestation on its own homepage. The footer carries a SOC 2 badge, an image at wp-content/uploads/2025/07/soc_2.svg that renders as a SOC 2 Type II mark, displayed beside the company's NVIDIA Inception badge.

What is missing is anything a procurement team can inspect behind that claim. The trust.cyata.ai subdomain resolves to an empty Notion shell with no controls list, and there is no downloadable report or self-serve trust portal, so a buyer can see the badge but cannot pull the audit period, scope, or auditor without a sales conversation. Sitting inside Check Point could improve that posture if the acquirer folds Cyata into its compliance program and supplies the artifacts an early independent had displayed but not yet made inspectable, though the public record does not yet document that coverage. [s1, s10]

Add analyzed competitors to compare them side by side with Cyata.

Cyata aims at the enterprise security team standing up AI agents inside core business processes. The company frames the gap as a new class of digital worker that executes code, queries sensitive databases, and triggers workflows while operating outside the identity controls applied to human employees or service accounts. The buyer is the CISO organization that has to govern agents it cannot fully see.

The sprawl-discovery use case fits organizations whose agent use is already operational, and Cyata publicly positions the product for buyers just beginning to explore AI as well as those with agents deployed. Cyata describes discovering orphaned, shadow, and overly privileged agents across SaaS, cloud, and identity infrastructure, which presumes a buyer with enough agent sprawl to have lost track of it. That use case fits larger enterprises more naturally than small teams running a single copilot.

Named demand is thin but real. The homepage carries one public case study, an eCapital reference naming its VP of Technology and Compliance, alongside practitioner endorsements from a Tenable cloud-security leader and an Everon CIO, beside a wall of logo images that no text confirms as paying customers. The open question is how many of those logos convert to references now that Check Point owns the roadmap.

Cyata builds its capability claim around seeing agents other tools miss. The product discovers AI agents by scanning SaaS, cloud, and identity infrastructure and analyzing behavioral signals such as tool usage, API activity, and MCP interactions, catching active and latent agents including orphaned, shadow, and overly privileged ones. Discovery is the entry point, with recording and least-privilege enforcement layered on top.

The team's research is the strongest external proof that the capability is grounded in real agent behavior. Cyata researchers disclosed LangGrinch, a critical langchain-core flaw tracked as CVE-2025-68664 with a CVSS score of 9.3 that SiliconANGLE reports can exfiltrate secrets and potentially reach remote code execution, and the company's own research posts describe further work on Anthropic's official MCP server. That is original adversarial work on the exact agent infrastructure the platform governs.

The public research record is the clearest non-product asset the fetched sources demonstrate, rather than a unique technical artifact. Discovery by behavioral scanning and least-privilege policy is engineering a funded rival can also build, and no named non-public dataset or third-party accuracy benchmark appears in fetched sources. What the record shows is original adversarial work published in public. It does not show buyers choosing Cyata for that work, or how long a rival would need to build a comparable research profile.

Cyata's go-to-market engine is research that points outward rather than at named customers. Its LangGrinch disclosure against langchain-core drew independent coverage from SiliconANGLE, and Cyata's own research posts describe further work on agent tooling such as Anthropic's official MCP server, building a profile larger than a seed-stage company would otherwise hold. That coverage made Cyata's research team worth following, an attention and credibility engine rather than evidenced buyer demand.

Commercial proof in the public record stays light. The homepage shows one named case study, the eCapital reference, plus a set of practitioner testimonials and a logo wall whose brands appear only as images, and TLV Partners led the $8.5 million seed, which is investor conviction rather than buyer proof. There is no published customer count or revenue figure in fetched sources.

The acquisition short-circuits the usual go-to-market path. Check Point announced the purchase about six and a half months after Cyata left stealth, and Globes reports the technology will be integrated into Check Point's AI Security platform. No fetched source states what drove the deal, so the headline external event is an acquirer adding Cyata's capability rather than a deep roster of references the company built itself.

Cyata publishes no pricing in fetched sources, so the charged unit and list price stay private. The homepage routes a buyer to a demo and contact flow rather than a self-serve signup, which now points to Check Point's contact-us page. That demo-led entry reads as a negotiated enterprise motion aimed at the larger organizations the discovery pitch targets, though no fetched page states a charged unit, list price, or negotiated terms, so the enterprise-pricing read is an inference rather than a sourced fact. The absence withholds the budget-anchoring signal some peers publish.

Neither the price nor the charged unit is publicly disclosed. No fetched page names seats, events, agent count, or any other value meter, so the basis on which the platform charges stays unknown.

The hidden-price posture fits the stage and the buyer. An early company selling a governance layer to enterprise security teams typically negotiates each deal. How Cyata is packaged, priced, and contracted under Check Point does not appear in fetched sources, which describe the deal and a planned technology integration without addressing the commercial motion.

Cyata delivers as software the customer connects to its own environment. The product scans SaaS, cloud, and identity infrastructure to discover agents, assigns dynamic identities, records what each agent does, and enforces least-privilege policy, all running against systems the customer already operates. No managed-service or analyst-staffed option appears in fetched sources, so delivery reads as customer-connected software rather than an operated service.

The fetched pages describe an observe-and-govern operational footprint without documenting where enforcement sits. Cyata says it assigns dynamic identities, enforces policy guardrails that include automatic shutdowns and predefined human interventions, and feeds its agent logs into SIEM, SOAR, and compliance systems, so it depends on the SaaS, cloud, and identity sources it reads from. No fetched page describes deployment effort, topology, published uptime, or support SLAs.

The acquisition reshapes the delivery question going forward. Globes reports Cyata's technology will be integrated into Check Point's AI Security platform, while hosting, support, and packaging under that owner stay public unknowns in fetched sources.

Cyata displays one security attestation for a product that inspects privileged agent activity. The homepage footer carries a SOC 2 badge, an image at wp-content/uploads/2025/07/soc_2.svg with empty alt text, beside an NVIDIA Inception partner badge and an OWASP membership tier badge that are program affiliations rather than security certifications. That is a single self-displayed mark, not a broad attestation set.

Nothing behind the badge is inspectable in fetched sources. Beyond the SOC 2 badge, the homepage FAQ says Cyata logs support frameworks such as PCI, SOC 2, and ISO 27001, which is framework-support language rather than a held certification. No displayed ISO 27001 certification, downloadable report, audit scope, or inspectable trust portal appears in the cited public sources. A buyer can see the SOC 2 claim, and the cited public pages do not expose the audit period, scope, or auditor behind it.

The product handles sensitive material, which raises the assurance bar the single badge does not clear. Because Cyata observes agent actions, API activity, and identity data across a customer's environment, a buyer should resolve data-handling, retention, and model terms in a formal review. Post-acquisition packaging may change, but fetched sources do not show Check Point trust coverage extended to Cyata.

Cyata markets itself as a control plane for agentic identity, but the product reads as a layer on top of the existing stack. It scans SaaS, cloud, and identity infrastructure to find and govern agents, depending on those source systems, and the fetched sources do not document agents authenticating through Cyata to function. The control-plane claim is grounded in the governance view, while the sources leave the credential path unresolved.

Whatever its exact position, the capability is absorbable by larger owners. Discovery and least-privilege enforcement for agents is adjacent to the identity platforms and security suites that already sit inside the enterprise, so the governance layer Cyata builds is ground a larger owner can contest by extending products the buyer already runs.

A platform vendor bought Cyata months after its public launch, and Globes reports the technology is slated for integration into Check Point's AI Security platform. The packaging and timing of the product under its new owner are not yet public.

Cyata's credibility comes from a founding team with both offensive-security craft and prior operating experience. The company was founded in 2024 by CEO Shahar Tal, who previously worked at Check Point and was a senior executive at Cellebrite, together with CTO Baruch Weitzman and VP of R&D Dror Roth, and the wider team includes alumni of Unit 8200. That is verifiable in-domain pedigree rather than LinkedIn-only background.

The research record is the team's strongest public signal. Cyata published a sustained stream of adversarial findings in its own domain, from LangGrinch in langchain-core to its research on Anthropic's official MCP server, a pattern of disclosures rather than a single covered event. SiliconANGLE covered the LangGrinch work independently.

The acquisition ties the team's standing to a commercial outcome. Check Point announced the purchase months after Cyata left stealth, and the acquirer is a company the CEO previously worked for, though no fetched source states what drove the deal. Depth below the principals stays the open question fetched sources do not settle.