Eve Security
Austin startup building "agent-in-the-loop" runtime policy enforcement that restricts what AI agents are allowed to do based on intent and data sensitivity.
Visit Website ↗ + Add to Compare Claim This CompanyOverview
Eve Security sells a runtime observability and policy-enforcement platform for agentic AI systems. Its product inspects the intent and data content of requests an AI agent is about to execute and can block or flag actions in real time, rather than only logging them after the fact. It supports API, MCP, and A2A agent-communication protocols and has built out discovery, enforcement, and automated remediation for environments including Databricks, Microsoft Copilot Studio, Amazon AgentCore, Amazon Bedrock, and Glean.
Founded by CEO Nadav Cornberg along with CRO Sharon Eilon and CTO Amit Eliav, the Austin-based company launched publicly in January 2025. Its differentiator, which it calls “Agent-in-the-Loop,” keeps a human-supervised policy layer embedded in agent workflows so that only genuinely risky actions get flagged, instead of generating broad alert noise the way session-level monitoring tools do.
The company has raised a seed round backed by Live Oak Ventures, Tau Ventures, and Run Ventures, with reported totals varying between roughly $4.5 million and $7.5 million across disclosures. Its most notable named reference customer is ZigiWave, an integration platform serving Fortune 500 clients including NASDAQ and Vodafone, which selected Eve Security to secure its agentic AI deployments.
Innovation Matrix Assessment
Went from January 2025 launch to integrations across five major agent platforms (Databricks, Copilot Studio, AgentCore, Bedrock, Glean) within roughly a year.
Lets security teams set intent-aware policies that restrict agent actions in real time, which is operationally useful as agentic AI deployments scale inside enterprises.
A real, named enterprise customer (ZigiWave) and multiple seed investors are genuine signals, but total funding is still small and reported figures are inconsistent across sources.
The "Agent-in-the-Loop" framing is a thoughtful refinement of policy enforcement, but it sits within an already crowded and fast-growing agent-security tooling space rather than redefining it.
Evidence rests on one named customer reference rather than independent testing or multiple disclosed deployments.
As enterprises hand more autonomous actions to AI agents, runtime policy enforcement for agent behavior will remain a necessary control layer.
Why CISOs Should Care
Provides a way to restrict what an AI agent can actually do at the moment of action, based on sensitivity and intent, rather than relying only on post-hoc logging.
What Makes It Different
Keeps a human-supervised 'agent-in-the-loop' policy layer embedded directly in agent workflows, evaluating intent and data exposure per action rather than per session.
The Matrix Verdict
57/100 — INCREMENTAL INNOVATOR
Eve Security is an Incremental Innovator: a credible, fast-moving entrant into AI agent runtime security with one strong reference customer, but still early-stage with funding and efficacy evidence too thin to score higher.
Editorial Note: Claims vs. Verified Findings
The ZigiWave customer relationship and platform integrations are corroborated by multiple outlets; exact funding totals conflict between sources ($4.5M vs $7.5M) and are reported here with that caveat rather than resolved in the company's favor.
Sources
Alternatives to Eve Security
Adaptive Security
AI-driven platform that simulates deepfake, voice, and multichannel social-engineering attacks to train and test organizations against next-generation phishing.
Quilr
Early-stage agentic AI security startup building a 'Service-as-Software' platform to guard against human-related breaches and secure AI agent…
Zenity
Governance and security platform for AI agents and low-code/no-code development, securing agent identity, permissions and behavior across the…
Gray Swan AI
Carnegie Mellon-linked startup building automated red-teaming and runtime guardrails that frontier AI labs use to stress-test their models…
Tenzai
An agentic AI penetration testing startup building autonomous 'AI hackers' to find and validate exploitable vulnerabilities at a…
Alice (formerly ActiveFence)
Israeli AI security company (rebranded from ActiveFence in January 2026) offering a lifecycle platform to test, guard, and…