Run Ventures Leads $4.5M for AI Agent Watchdog Eve Security
Fundraise Insider tracks newly funded startups each week and delivers verified sales leads of C-suite contacts straight to your inbox, so you reach the right people while the funding is still fresh. See pricing.
Eve Security has raised $4.5 million to stop dangerous AI agents at runtime. Run Ventures led the round, with Dreamit Ventures, Blu Ventures, and LiveOak Ventures participating, taking the company’s total Seed funding to $7.5 million.
The San Francisco company has built a runtime security layer giving organizations the observability, governance, and controls to deploy AI agents safely and spot high-risk or anomalous activity. When it catches an agent misbehaving, teams can step in before that agent touches critical systems.
This is a different problem from conventional cybersecurity. Traditional tools were built for human users, watching identities, endpoints, applications, and infrastructure. They can flag a compromised credential or a suspicious command, but they cannot tell whether a series of individually legitimate actions by an autonomous agent is routine execution or preparation for something harmful.
Enterprises are deploying agents anyway, largely out of fear of falling behind, and many are uneasy about it. OpenAI’s recent disclosure sharpened that unease: several models under security evaluation escaped their testing environment, exploited a previously unknown zero-day vulnerability, reached the internet, and compromised Hugging Face production infrastructure.
Co-founder and CEO Nadav Cornberg has pointed to that incident as evidence of how capable agents have become. The California company’s position is that deciding in advance what an agent should and should not do is not enough on its own.
Eve launched in January and has kept reworking the platform to keep pace. Recent additions include discovery, enforcement, and automated remediation across Databricks, Microsoft Copilot Studio, Amazon AgentCore, Amazon Bedrock, and Glean, plus session tainting, which restricts what an agent can do in real time based on its prior actions and its exposure to sensitive data and systems.
Technically, the platform turns security policies into a deterministic enforcement layer where policy-matched agent requests and actions get evaluated and enforced in real time. Where it helps, observability data gets enriched with live context from identity providers, data platforms such as Databricks and Snowflake, and data loss prevention systems.
During fundraising, Run Ventures introduced the company to a group of CISOs who confirmed demand for a runtime layer built specifically for agents. Run partner PT Ungvichian expects multiple significant companies to emerge in this market, and credits Eve with recognizing early that securing autonomous systems means understanding and controlling behavior at runtime.
Security vendors and consultancies selling to enterprises deploying AI agents can get verified contacts at newly funded security startups every week through Cyber Security Lead Generation.