Vero Security is launching a free community for the people tasked with keeping enterprise AI agents under control, pairing an educational roadmap with a commercial platform designed to turn security policies into restrictions that software can enforce.
Announced November 18, AISecurityEngineering.org targets professionals learning to secure systems that can read corporate information, connect to applications and act on employees’ behalf. Vero’s accompanying product pitch addresses the operational burden: security teams decide what agents should be allowed to do, and its platform implements and maintains the corresponding controls.
The company also announced $11 million in seed funding, led by TLV Partners with participation from GTM Capital, Maple Capital and angel investors. Founded in 2025, Vero says multiple Fortune 500 companies already use its software, although it does not identify them in the announcement.
The combined initiative makes a broader argument: securing AI is becoming a job with its own skills and recurring responsibilities, even when the person doing it started elsewhere in the security organization.
“That’s a new profession, not a side gig of something that has existed for many years,” cofounder and CEO Elad Tessler told VentureBeat in an interview, describing customers shifting their attention from endpoint and cloud security toward AI infrastructure.
A free roadmap, with credentials still to come
The community’s most concrete initial offering is a learning roadmap spanning 259 topics across nine available tracks. Those cover security fundamentals, how agents operate, threats, asset inventories, configuration, identity, runtime protections, monitoring and incident response. Two additional tracks, covering program management and agent red-teaming, are marked as forthcoming.
The roadmap is accessible without registration, and membership is free. That gives security engineers and managers a starting point for understanding both agent behavior and the administrative work of documenting what their organizations use and what those systems can access.
There is a distinction between what Vero announces and what visitors can use immediately. The release describes guides, workshops and an ASE credential track. The website marks member-sharing and meetup features as forthcoming and offers notifications when credentials open. It does not establish that a certification is available today.
Vero explicitly identifies itself as the site’s builder and funder. Its homepage says the curriculum excludes product promotion and that sales staff do not receive members’ names. This is a vendor-backed educational initiative, rather than an independent standards organization.
Turning a security decision into working restrictions
Vero’s commercial platform addresses a related problem: stating a rule is easier than enforcing it across different agents, applications and permissions.
The company describes an unnamed Fortune 1000 customer that wanted only developers to connect agents to corporate repositories or authorize code execution. Vero says it converted that decision into more than 15 controls within minutes, spanning coding agents, identities, skills, execution permissions and Model Context Protocol servers, which connect AI systems to external tools and data.
Its AI Runtime Governor checks supported prompts, tool invocations and agent operations as they occur. Advertised protections include preventing agents from opening sensitive credential files, blocking prohibited actions and limiting unauthorized behavior triggered by malicious instructions or skills.
The platform also tracks changes in AI applications and configurations. Security teams set the boundaries for automatic adjustments; decisions involving new risks return to a person for approval.
“AI touches everything in the organization, and we integrate to those services,” Tessler said, explaining why Vero connects to existing web, cloud, endpoint and business-application infrastructure.
The website lists connections to products including CrowdStrike, Microsoft Intune, Okta, GitHub and Slack. Employees need not install software themselves, according to Vero, but that does not mean every deployment is software-free: browser protection can involve an extension distributed through device-management tools.
AI interprets policy; technical controls enforce it
Vero’s architecture draws a distinction between using AI to understand a security goal and relying on technical restrictions to enforce it. Its platform documentation describes AI interpreting environmental changes and preparing updates, with explicit permission rules governing what agents can actually do.
“So we have foundational models, we have our own fine-tuned models, but also not only LLMs,” Tessler said. He also identified decision models such as Jev as part of the company’s approach, without providing a complete model inventory or performance breakdown.
That distinction matters for evaluating the product. An agent’s access should depend on enforceable restrictions, while the process that generates those restrictions still needs validation. The supplied materials do not provide independent measurements showing how reliably Vero translates policies or how often its protections block legitimate work.
Competitive strengths and gaps: automation is the differentiator to test
Vero enters a market where large security vendors already advertise overlapping protections. Its strongest proposed distinction is the ongoing work of translating security intent into controls, coordinating changes and maintaining them across a changing environment. Runtime blocking alone does not distinguish it.
Palo Alto Networks’ Prisma AIRS Agent Security advertises agent discovery, identity and permission governance, supply-chain inspection, adversarial testing and runtime enforcement. Those capabilities overlap substantially with Vero’s inventory and control story. Palo Alto’s broader documented testing and scanning portfolio gives buyers additional capabilities to evaluate alongside everyday policy administration; it does not, by itself, establish better protection.
Cisco AI Defense likewise combines discovery, component scanning, automated attack simulations and runtime safeguards. Its Policy Studio uses a conversational agent to help refine custom guardrails and assess their accuracy. That makes even AI-assisted policy creation an area of competition, rather than an exclusive Vero capability. Cisco also documents a deployment option that keeps inference traffic within customers’ private infrastructure.
SentinelOne’s Singularity Agentic Detection and Response, announced in early access in October, connects agent governance to its existing behavioral detection and response platform. It advertises tracing agent activity back to user requests and stopping individual agents during execution. For existing customers, the integration offers a potential advantage in investigation and response workflows.
Vero’s advertised ability to operate through multiple vendors’ existing tools could appeal to enterprises with mixed infrastructure. Its emphasis on reducing recurring administrative work could also matter to understaffed teams. Both are propositions to test in deployment, rather than demonstrated advantages over the alternatives.
The principal evidence gap is practical: Vero supplies no named customer references, independent comparative tests, false-positive rates or enforcement-latency measurements in these materials. Buyers also need to distinguish integrations that discover activity from those that can block it. A product logo on an integration page does not establish equivalent enforcement across every supported application.
Annual contracts, undisclosed prices and deployment choices
Vero sells its platform commercially, with no open-source license identified in the supplied materials. Prospective customers can request a demonstration; the announcement does not describe self-service access or geographic restrictions.
“So it’s standard pricing per year, even though there is a lot of AI in the backend,” Tessler said. He explained that pricing depends on the number of users and AI assets, with predictable annual contracts. Vero did not disclose dollar amounts.
There is a meaningful packaging distinction with Prisma AIRS licensing, which uses software credits and calculates API-interception consumption in monthly token volumes. That documentation does not provide a directly comparable dollar quote. Neither Cisco’s cited overview nor SentinelOne’s launch announcement specifies pricing, leaving insufficient information to rank these offerings by cost.
Vero also offers deployment flexibility. “We also have a hybrid deployment and local deployment,” Tessler said. Some models run on customer devices; other processing uses its backend, including AWS infrastructure. The interview does not establish a universal zero-retention policy or specify all supported regions and data-handling terms.
Vero plans to use the funding to broaden its technology coverage and expand its U.S. commercial presence. For enterprises evaluating the community and product, the immediate opportunities are different: the roadmap offers a free starting point for building expertise, while the platform requires proof that automating policy maintenance reduces work without introducing new security gaps.
