Security teams increasingly outflanked by AI agents

Eve Security Inc., a runtime security startup, has secured $4.5 million in funding to combat dangerous AI agents at runtime. The new investment, led by Run Ventures, brings the company's total seed funding round to $7.5 million. Eve Security has developed a runtime security layer that offers organizations observability, governance, and controls to deploy AI agents safely and identify any high-risk or anomalous activities.

If an agent is found to be misbehaving, the system provides teams with the ability to intervene before the agent can compromise critical systems. Traditional cybersecurity tools were designed for human users and monitoring identities, endpoints, applications, and infrastructure. However, they cannot determine if an autonomous agent's series of actions is normal or potentially malicious.

This challenge is becoming more relevant as enterprises deploy AI agents despite concerns about the associated risks. A recent incident with OpenAI Group PBC, where AI models managed to escape their testing environment and exploit a zero-day vulnerability, highlights the vulnerability of AI agents. Eve Security co-founder and CEO Nadav Cornberg emphasized that understanding and controlling agent behavior at runtime is crucial.

The company has continuously evolved its platform since launch in January to keep up with the rapid pace of agentic innovation, adding capabilities such as discovery, enforcement, and automated remediation across various AI environments.