Vero Security Raises $11M Seed to Police AI Agents at Runtime
TLV Partners leads $11M seed in Vero Security, whose AI Runtime Governor enforces policy on prompts, tool calls and agent actions while agents work.
By Amara Osei
4 min read

What's News
- Vero Security raised an $11 million seed round led by TLV Partners, with GTM Capital, Maple Capital and angels participating.
- Founded in 2025, Vero says multiple Fortune 500 companies already use its platform.
- One Fortune 1000 customer's single policy became more than 15 controls within minutes.
- Founders Elad Tessler and Daniel Meirov co-led a Unit 8200 cyber operations unit of more than 100 people.
- Vero also launched AISecurityEngineering.org with field guides, workshops and an ASE credential track.
Vero Security has raised $11 million in seed funding to turn enterprise AI security policies into controls that operate while employees' AI tools and agents do their work. TLV Partners led the round, with GTM Capital, Maple Capital, and angel investors participating.
Founded in 2025, Vero says multiple Fortune 500 companies already use its platform. The startup plans to spend the capital on product development, broader coverage of enterprise AI technologies and security workflows, and expansion of its U.S. commercial presence. It has offices in New York and Tel Aviv.
The company targets a problem that grows with agent usefulness: software that reads company files, connects to repositories, invokes tools, and executes code needs boundaries that stay effective as those capabilities change.
What does Vero actually enforce?
Vero calls its market AI Security Engineering: the continuous work of discovering how AI is used, assessing the risks, and keeping protections current. Its pitch is to take on that operational workload rather than hand security teams another list of findings.
Security professionals define the policy and decide which changes the platform may implement automatically. Vero translates approved decisions into controls, coordinates fixes with the relevant owners, verifies completion, and adjusts protections as tools and permissions evolve. Changes requiring a fresh risk decision go back to a person, with supporting analysis and a recommended action.
One Fortune 1000 customer, Vero says, started with a policy allowing only developers to connect agents to corporate code repositories or execute code. The platform translated that single statement into more than 15 controls — spanning coding agents, identities, Model Context Protocol servers, skills, and runtime actions — within minutes.
The example shows the engineering work hidden inside a simple policy. Enforcing who may use an agent is only part of the task; the connected tools, credentials, and operations must also reflect that decision.
Why runtime enforcement?
At the platform's center sits Vero's AI Runtime Governor. The company says it inspects supported prompts, tool calls, and agent actions, applying policy when an operation is attempted rather than waiting for a later manual review.
Potential interventions include preventing a coding agent from reading unrelated secrets, blocking actions outside an approved policy, and stopping malicious skills or prompt injection from triggering unauthorized behavior. Prompt injection occurs when material an AI system encounters contains instructions intended to redirect its behavior. For a text-drafting assistant, the consequence may be an unreliable answer. An agent with repository access or command-execution permission can turn the same manipulation into an operation against a real system.
That makes runtime enforcement distinct from a policy written into an agent's instructions. A rule asking software to avoid a sensitive resource expresses intended behavior. An enforcement mechanism must inspect the attempted access and apply the boundary along the execution path it supports.
Are configuration checks enough?
Vero's website describes four connected functions: inventory, posture analysis, behavior detection, and enforcement. Inventory identifies agents and their owners; posture analysis examines configuration and reachable credentials; behavior detection examines session activity; enforcement applies the policy.
Its illustrated workflow follows an agent whose configuration exposes production credentials and whose approved skill later receives a malicious update. Vero says it addressed the configuration problem, then blocked the harmful call when the changed skill tried to act.
The distinction matters because an initial approval is a snapshot. A tool update, new connection, or added permission can change risk after deployment. MCP, one of the connection mechanisms involved, lets AI applications interact with external tools and data sources; skills supply reusable instructions. Together with agent credentials, these components determine what an agent can actually do.
Vero says it operates through the endpoint, identity, AI, and security tools an enterprise already uses, without requiring employees to install something new. Its enforcement claims apply to supported prompts, tools, and actions — so the practical scope of a deployment follows the integrations and execution paths it covers.
Who is behind the bet?
Elad Tessler, CEO, previously led product at Astrix Security. Daniel Meirov, CTO, was Sightfull's first employee and led its R&D and product work. The founders met in Unit 8200, where they co-led a cyber operations unit of more than 100 people, according to the announcement.
GTM Capital says it invested after a structured evaluation involving its community of Fortune 500 CIOs and CISOs, with those executives working directly with Vero before the firm committed.
The round lands as runtime controls become a bigger part of enterprise AI security; Sysdig's AI Defense also observes agent behavior and enforces boundaries during execution. Voro's pitch emphasizes the ongoing operating work: implementing decisions, verifying remediation, and keeping controls aligned with policy.
Alongside the funding, Vero launched AISecurityEngineering.org, an open community and knowledge center with field guides, workshops, and an ASE credential track, aimed at professionals building AI security practices alongside other duties.
The broader test, as enterprises hand agents more access, is whether platforms like Vero can consistently carry a security team's decisions through to the actions those agents attempt.
Source: unite.ai
More from Amara Osei
Show full bio
Senior reporter covering companies, deals and strategy in entrepreneurship, startups and venture funding at Business Bearings.
2,260 articles