Bloom Security

**Product and the concrete problem it solves.** Bloom Security is aimed at a specific change in enterprise risk: the endpoint is no longer just an operating system, a set of managed applications, and a known collection of executable files. Employees now install coding assistants, browser extensions, IDE plugins, open-source packages, local models, MCP servers, autonomous agents, and other software that can read local files, access credentials, call external services, or transmit company data. Traditional endpoint detection and response is valuable for identifying known malicious behavior, but it does not necessarily answer the operational question Bloom is targeting: is this tool safe on this particular endpoint, for this particular user, with this user's access and the other software already present? Bloom's public product thesis is that security teams need continuous visibility and active control over the whole endpoint ecosystem, including AI tools and non-binary software. The platform therefore presents itself as a complete endpoint solution with live inventory, contextual risk analysis, granular remediation, active prevention, and AI guardrails. The practical outcome is intended to be a smaller attack surface, fewer unmanaged tools, and a way for organizations to let employees use useful AI without accepting uncontrolled access to sensitive systems.

**Core technology and how it works.** Bloom describes a device and control-plane architecture rather than a single detection model. Its live-inventory layer discovers the operating system, local data, installed applications, extensions, AI tools, code libraries, agents, and related components, then monitors changes across the fleet. Its contextual analysis combines marketplace intelligence, static analysis, and behavioral sandboxing to assess how a tool was built, what it can access, where it came from, and how it behaves in the local environment. That context is important because a package that is harmless in one workstation may be dangerous on a developer machine with production credentials or on an administrator's laptop. Bloom also claims enforcement across ecosystems such as npm, the Chrome Web Store, and Open VSX, allowing security teams to apply policy to the points where packages and extensions enter the environment. The AI Guardrails capability is intended to define what agents can access, execute, and transmit, while automatically reducing over-permissioned MCP configurations to safer defaults. The company has not publicly disclosed model architectures, telemetry volumes, endpoint-agent internals, policy language, or independent performance benchmarks, so the technical description should be treated as a product-level account rather than a verified claim of proprietary machine-learning superiority.

**Market, customers, and go-to-market.** Bloom is selling into the overlap of endpoint security, software supply-chain security, AI governance, and enterprise application control. The immediate buyer is likely a CISO or endpoint-security leader whose organization has accumulated overlapping EDR, MDM, application-control, browser-security, and AI-governance tools but still cannot establish what is actually running on employee machines. The wedge is timely because security teams are being asked to approve AI adoption while employees and developers can download new tools faster than a conventional review process can evaluate them. Bloom's platform can be positioned as an additive control plane for discovery and prevention, but long-term enterprise value depends on whether it consolidates or displaces existing agents rather than becoming another endpoint sensor. The company has public addresses in Tel Aviv and Wilmington, Delaware, and its launch materials emphasize continued research and product-team expansion in Israel. The public record does not name commercial customers, contract values, annual recurring revenue, renewal rates, or deployment counts. The reference quote on the official website is attributed to Bradley Schaufenbuel, a VP and CISO at an unnamed HCM market leader; that is evidence of buyer resonance, not proof of a disclosed customer relationship. Initial go-to-market is therefore best understood as enterprise security sales supported by a technically credible founding team and investor networks, with customer scale still to be established.

**Traction, funding, and third-party validation.** Bloom announced its launch from stealth on July 30, 2026 with a $20 million Seed round led by Glilot Capital Partners, with participation from Ten Eleven Ventures, Okta Ventures, and Runtime Ventures. The round also included cybersecurity angels associated with Dig Security, Demisto, Snyk, and Talon, giving the company a syndicate with relevant experience in cloud data security, security operations, software supply chain, and enterprise-browser protection. Axios reported the financing and identified CEO Itay Keren, while CTech independently described the company as an Israeli startup founded in 2025 by Keren, Ofir Balassiano, and Itay Frishman. The official investor announcement confirms the financing, Israeli launch location, and the company's focus on the AI-native endpoint. Startup Nation Finder records a November 2025 founding date, a 1-10 employee range, a released product, and an Israeli company-registration entry, although those directory details should be treated as a snapshot rather than a current audited cap table or headcount report. Bloom's own site gives a more useful product-level validation by documenting specific controls and by showing a live hiring posture in Israel. What remains missing is equally important: no independently measured detection or prevention rate, no named customer list, no formal certification publicly tied to the platform, no patent portfolio disclosed, and no evidence yet of a large production deployment. The $20 million round is a strong ecosystem signal, but it is not itself evidence that the product has won a mature market.

**Founders and team background.** Bloom's three founders bring unusually direct experience for the problem they are addressing. CEO Itay Keren previously held engineering and sales-engineering leadership roles at Palo Alto Networks and at Dig Security and Demisto, both of which were acquired by Palo Alto Networks. CPO Ofir Balassiano led the Cortex Cloud Posture Security research group at Palo Alto Networks, previously led security research at Dig Security and XM Cyber, and began his career in the IDF's Mamram technical unit. CTO Itay Frishman is described by Bloom as an engineering leader who built core AI security posture-management and data-security posture-management solutions at Palo Alto Networks and Dig Security, with prior cybersecurity R&D leadership in the IDF's Unit 81. This background matters because the product sits at the boundary between endpoint telemetry, data access, application behavior, and security policy rather than in a single conventional antivirus category. The founders have seen how large vendors integrate acquired technology into broad platforms, how enterprise buyers evaluate risk controls, and how difficult it is to turn research concepts into production security products. At the same time, the team is still small and the public record does not disclose its broader engineering bench, advisory structure, or operational leadership. The central team diligence question is whether experience inside a major platform translates into the speed, product focus, and support capacity required to build a new endpoint control plane from a Seed-stage company.

**Competitive dynamics and edge.** Bloom enters a market with powerful distribution and several adjacent Israeli challengers. CrowdStrike and SentinelOne already own endpoint agents, telemetry, and security-console relationships at many target accounts; Microsoft Defender for Endpoint and Intune can bundle application-control and device-management features into existing enterprise agreements; Tanium competes on real-time asset inventory and software control; and Island competes for the browser and distributed-workforce security budget. Koi Security, although a distinct company, is the closest conceptual comparison in the local database because it also addresses AI agents, MCP servers, extensions, packages, and endpoint software that traditional tools miss. Glow is another distinct local comparison, with a broader prevention-first endpoint platform and a much larger disclosed capitalization. Bloom's plausible edge is the combination of four elements: a live inventory that treats agents and extensions as first-class assets; contextual, per-endpoint risk judgment instead of static allowlists; prevention at the software-supply-chain entry point; and founders who understand the enterprise-security acquisition and platform landscape from inside Palo Alto Networks. The product could be more useful than a narrow AI scanner because it connects discovery, policy, remediation, and prevention. None of those advantages is yet a proven moat. Incumbents can extend their agents, dedicated competitors can specialize more deeply in AI components, and the endpoint agent itself can become a procurement objection. Bloom must show that its context produces fewer false positives, catches meaningful risks earlier, and reduces total security-tool cost rather than adding another console.

**Defense, security, and resilience dual-use relevance.** Bloom's dual-use relevance is credible because the core problem is not limited to ordinary commercial laptops. Defense contractors, government agencies, hospitals, utilities, financial institutions, and other critical-infrastructure operators increasingly depend on developer workstations and administrative endpoints where an unapproved extension, malicious package, over-permissioned AI agent, or compromised update can become a path to sensitive data or operational disruption. A continuously validated inventory, contextual software-risk engine, package and extension enforcement, and agent permission controls could help those organizations harden the software perimeter around systems that cannot rely solely on network segmentation or periodic compliance audits. The supply-chain angle is particularly relevant to resilience: blocking a malicious component before installation can be less costly than detecting it after it has executed, while identifying a silently degraded security agent can close a blind spot that conventional consoles may not report accurately. Bloom's Israeli provenance also places it within a security ecosystem with substantial experience in endpoint, cloud, industrial, and intelligence environments. The connection remains an adjacency, not a fielded defense capability. Public sources reviewed here do not establish an Israeli Ministry of Defense contract, classified deployment, defense-prime customer, government authorization, air-gapped operating mode, or use in an operational mission. For public-sector adoption, diligence would need to test sovereign or self-hosted deployment, data residency, offline policy behavior, tamper resistance, audit retention, identity integration, export controls, and performance under degraded connectivity. Bloom is therefore dual-use through its defensive control technology and resilience applications, not because it has disclosed a military-specific product.

**Growth stage, trajectory, and key diligence risks.** Bloom is classified as early: it was founded in 2025, launched publicly in July 2026, raised a substantial Seed round, and has a released platform, but the public record does not establish revenue, retention, customer count, or scaled deployment. Its trajectory depends on whether AI changes the endpoint-security buying center faster than incumbents can absorb the category. A successful path would make Bloom the control layer for everything a user or agent installs, then expand from visibility into enforcement, remediation, AI governance, and software-supply-chain assurance. The main risks are (1) incumbent bundling by Microsoft, CrowdStrike, and SentinelOne; (2) agent fatigue and long enterprise replacement cycles; (3) false positives from autonomous blocking or permission reduction that interrupt legitimate developer or business work; (4) false negatives in rapidly changing packages, extensions, and agent ecosystems; (5) the cost and privacy burden of collecting enough endpoint context to make reliable per-user decisions; (6) competition from narrower AI-security vendors and browser or identity platforms; and (7) insufficient public evidence that the product works outside demonstrations. The $20 million round gives Bloom time to build, but it also raises the milestone bar for the next financing. Useful diligence milestones include named production customers, independently measured prevention and remediation outcomes, a clear endpoint-agent architecture, security attestations, evidence of displacement or consolidation, and a public-sector or defense-industrial evaluation. Until those appear, Bloom is a high-quality early Israeli cyber entry with strong team signals and strategic relevance, but with commercial proof and technical differentiation still to be earned.