Cisco Systems is in talks to acquire Astrix Security, a Tel Aviv-based startup that helps enterprises discover and secure their deployed AI agents, for between $250 million and $350 million. The Information first reported the negotiations on April 10, with SiliconAngle and Digital Today confirming the deal range.
The price tag is roughly three times the total funding Astrix has raised to date. Investors in the five-year-old company include Workday and the Anthology Fund, an investment vehicle backed by Anthropic.
What Astrix Does
Astrix’s platform automatically inventories every AI agent deployed in a company’s network, identifies the tools each agent uses, and detects the MCP servers and non-human identities they rely on to authenticate with external applications. The software then scans those assets for vulnerabilities: agents with excessive permissions, internal agents exposed to the public web, coding assistants with delete access to repositories they should only read.
The platform includes just-in-time (JIT) access controls that expire agent credentials after configurable time windows, according to SiliconAngle. A JIT policy might specify that a database agent’s credentials expire after 10 minutes. Expired credentials are useless to attackers even if compromised.
Astrix also monitors for anomalous agent behavior, such as attempts to bulk-download sensitive records, and can revoke malicious agents’ access without manual intervention. Findings sync to third-party cybersecurity tools for automated breach investigation.
On the provisioning side, administrators configure security policies in advance, letting developers deploy new agents on a self-service basis without security team bottlenecks.
Two Acquisitions in Two Days
The Astrix talks came just one day after Cisco announced its agreement to acquire Galileo Technologies, a startup selling hallucination firewalls that protect AI models from malicious prompts and prevent sensitive data leakage in responses. Cisco plans to integrate Galileo into Splunk’s observability stack.
Combined with its 2024 acquisition of Robust Intelligence (which focuses on preventing security vulnerabilities in AI applications), Cisco is assembling a three-layer agent security stack: model protection (Robust Intelligence), output monitoring (Galileo), and agent governance (Astrix).
The Enterprise Bet
The acquisition spree reflects a broader shift in how enterprises are approaching AI agents. Discovery, identity management, and runtime security for autonomous agents are moving from research topics to procurement line items. Cisco’s move signals that for the networking giant, agent security is now a core subscription revenue play, not an edge-case product feature.
The talks are ongoing and may not result in a completed deal, according to The Information.