Eve Security raises USD $7.5 million for AI agent tools

Eve Security has extended its seed round to USD $7.5 million, with the latest USD $4.5 million investment led by Run Ventures.

The Austin-based company sells runtime security tools for AI agents, a segment focused on monitoring and controlling what autonomous software does after it has been given access to corporate systems and data.

The round also included Dreamit Ventures and Blu Ventures, while existing backer LiveOak Ventures participated again. The new funding will mainly support go-to-market expansion and revenue growth.

The fundraising comes as security concerns around autonomous AI systems draw more attention from large organisations. Eve pointed to recent disclosures from a cybersecurity evaluation in which AI models escaped a testing environment, reached the internet through a previously unknown zero-day, and then compromised production infrastructure at Hugging Face.

Nadav Cornberg, Co-Founder and Chief Executive Officer of Eve Security, said the incidents reflect a broader shift in how companies need to think about AI security.

"Agents are becoming extraordinarily capable, and enterprises are giving them access to increasingly valuable systems. The OpenAI incident showed what can happen when an agent's pursuit of an objective takes it somewhere its creators never intended," said Nadav Cornberg, Co-Founder and Chief Executive Officer of Eve Security.

"You cannot secure that world simply by deciding in advance what an agent should and shouldn't do. You have to understand what it is doing, why it is doing it, and have the ability to intervene while it is happening," he added.

Runtime focus

Eve argues that AI agents create a different problem from traditional cyber threats because a system may carry out a chain of individually legitimate actions that still leads to a harmful outcome. That differs from established security tools, which are generally designed to detect compromised credentials, suspicious users, infected endpoints, or unauthorised access attempts.

Its platform is designed to give companies visibility into agent behaviour while tasks are underway, and to let them question or stop activity judged to be high risk before it reaches sensitive systems. According to Eve, that approach has gained traction with Chief Information Security Officers during customer discussions.

Run Ventures said the investment process included introductions to a range of security leaders, whose feedback helped test demand for a separate runtime layer for AI agents.

"AI runtime security is developing into a significant new security category, and we believe there will be multiple important companies built in this market," said PT Ungvichian of Run Ventures.

"Eve recognized early that securing autonomous systems requires understanding and controlling behavior at runtime, and the company is exceptionally well positioned to capitalize on that shift," Ungvichian added.

Cornberg said investor interest alone was not enough and that the company had been looking for clearer evidence from buyers.

"We had high conviction in the technical thesis. What we wanted next was market proof," he said.

"We're now seeing that pull directly from CISOs and customers. This round lets us put significantly more resources behind turning that early demand into a repeatable business," Cornberg added.

Product expansion

Eve began its go-to-market push in January and is now seeing customers expand their use of the platform. It described that as early evidence that runtime agent security could become a standard enterprise requirement rather than a narrow tool bought for a single use case.

The company is also expanding the systems its software can cover as businesses deploy more complex agent-based workflows. New additions include session tainting, a method that adjusts or restricts what an agent can do based on its exposure to sensitive data and its earlier actions in the same session.

Eve has also expanded discovery, enforcement, and automated remediation across Databricks, Glean, Microsoft Copilot Studio, Amazon AgentCore, and Amazon Bedrock. It added that more than 85 per cent of policy-matched requests can now be evaluated and enforced through deterministic rules, while more complex decisions are supplemented by data from identity providers, data loss prevention tools, and data platforms including Databricks and Snowflake.

The emergence of specialist suppliers such as Eve shows how quickly the security industry is adapting to the spread of AI agents in corporate environments. As companies give these systems more freedom to interact with applications, data stores, and infrastructure, investors and security teams are increasingly treating oversight of autonomous software as a distinct operational issue rather than an extension of existing application security.

Eve said its customers are already increasing usage of the platform as they move from initial testing to broader deployment.