Cisco Systems (NASDAQ: CSCO) is in advanced acquisition talks to purchase Astrix Security, a Tel Aviv-based startup that monitors AI agents and non-human identities inside enterprise networks, for between $250 million and $350 million, according to reports from April 2026. The deal targets a security gap that has widened sharply as companies deploy autonomous AI agents at scale — systems that act on behalf of humans, hold privileged credentials, and leave minimal audit trails.
What Astrix Security Actually Does
Astrix Security’s core product monitors non-human identities (NHIs) — a category that includes AI agents, OAuth applications, API keys, and service accounts operating inside enterprise environments. The software maps every NHI in a company’s environment, tracks what each one accesses, and flags anomalous or over-privileged behavior across cloud platforms including AWS, Google Cloud, Microsoft 365, and GitHub.
Founded in 2021 by Alon Jackson (CEO) and Idan Gour (CTO) — both veterans of Israel’s Unit 8200 intelligence unit — Astrix has raised approximately $65 million across two funding rounds. The Series B, a $45 million raise led by Bessemer Venture Partners, closed in 2024. Customers span financial services, healthcare, and technology sectors, where an AI agent with unchecked access to production databases is not a hypothetical threat.
The differentiation from traditional identity and access management (IAM) tools is structural. A human employee authenticates with a password and MFA. An AI agent authenticates via API key or OAuth token, runs continuously, and may never trigger a traditional security alert even while querying sensitive systems thousands of times per hour. Standard IAM tooling has no model for that behavior.
The Agent Security Gap Cisco Is Buying Its Way Into
Enterprise AI adoption has created a new attack surface that existing security tools were not designed to cover. Gartner projected in late 2025 that by 2027, more than 50% of enterprise security incidents will involve non-human identities — a figure that would have been nearly impossible to reach before the AI agent wave of 2024–2026.
The problem is structural. When a company deploys an AI agent to handle customer service, process invoices, or monitor infrastructure, that agent typically receives credentials that persist indefinitely — often with broader permissions than the task requires. Security teams at most enterprises lack visibility into how many agents are running, what they can access, and whether any have been compromised.
Astrix addresses this with continuous monitoring rather than one-time audits. The platform ingests data from identity providers and cloud platforms, builds a behavioral baseline for each non-human identity, and surfaces deviations in real time. The pitch is direct: you cannot secure what you cannot see.
Why Cisco Wants This Now
Cisco’s security portfolio — which includes Duo Security (acquired 2018 for $2.35 billion), Cisco Umbrella, and most significantly Splunk (acquired 2024 for $28 billion) — is increasingly focused on the intersection of AI and enterprise risk. Adding Astrix would give Cisco a dedicated product line for AI agent governance, a category where no incumbent currently holds a dominant position.
The timing is deliberate. AI agent deployment accelerated sharply in 2025, with major platforms from Salesforce (Agentforce), ServiceNow, and Microsoft all shipping agentic capabilities into existing enterprise contracts. Each deployment creates new non-human identities operating inside corporate networks. Cisco’s enterprise customers are already managing this problem; Astrix gives Cisco something concrete to sell them.
Competitive pressure reinforces the urgency. CrowdStrike, Palo Alto Networks, and Wiz — valued at $32 billion after its 2024 funding round — have all signaled active interest in the NHI security space. Cisco cannot afford to let a cloud security rival define this category first.
The 0 Million Price Tag in Context
At the reported high end, $350 million represents roughly 5–7x Astrix’s last reported ARR, which sources place in the $50–70 million range. That is a moderate multiple by current enterprise security standards — well below the 15–20x ARR multiples that security acquisitions commanded at the 2021–2022 peak.
For context: Cisco paid $28 billion for Splunk, approximately 7x ARR at acquisition. The Astrix deal, if it closes at $350 million, would represent a disciplined bet on a category-defining startup before the market fully prices in the AI agent security premium. If enterprise AI agent deployments double again in 2026 — consistent with current adoption curves — the same asset would command significantly more.
The acquisition also fits CEO Chuck Robbins’s stated strategy of building AI security capabilities through targeted M&A rather than internal development. Building Astrix’s detection and integration layer internally would likely take Cisco two to three years and hundreds of engineers. The acquisition buys market position now.
Who Else Is Competing in Agent Security
Astrix is not alone in the NHI security space. Active competitors include Clutch Security, Entro Security, and Oasis Security, all of which raised significant funding in 2024–2025. The category attracted more than $500 million in venture investment over that period, according to PitchBook data — making it one of the fastest-funded emerging security verticals on record.
What distinguishes Astrix is the breadth of its integration layer and its focus on AI-specific agent behavior patterns. Most competitors began from the API key management problem and expanded toward agents; Astrix built its detection models with agentic behavior in mind from the start.
A Cisco-owned Astrix with distribution across Cisco’s 90,000+ enterprise customers changes the competitive dynamics for every other NHI vendor. As earlier AI acquisition cycles have demonstrated, the window for independent operation in a newly validated category closes quickly once an enterprise giant stakes its claim. Expect Clutch, Entro, and Oasis to either raise at elevated valuations or attract their own acquirers within 12–18 months.
What This Means for Enterprise AI Deployments
Most companies running AI agents in 2026 are doing so with inadequate visibility into agent behavior. A 2025 survey by the Cloud Security Alliance found that 67% of enterprises had deployed AI agents in production environments, but only 23% had implemented any form of continuous monitoring for those agents’ activity. That 44-point gap is the market Astrix sells into.
The risks of insufficiently secured AI agents are not abstract. Agents operating with persistent credentials and broad access permissions are prime targets for credential theft and supply chain compromise. An attacker who acquires a legitimate agent’s API key gains access to everything that agent can reach — often without triggering a single traditional security alert.
Enterprises that rely on autonomous AI systems for critical workflows should read this acquisition as a market signal: agent security is becoming a compliance and governance requirement. The debate about AI autonomy and human oversight has largely been philosophical — the Astrix category is where that debate becomes operational, encoded in access policies and audited in incident reports.
MegaOne AI tracks 139+ AI tools across 17 categories, and the NHI security segment is among the fastest-moving in the current cycle. The pace of enterprise agent deployments has consistently outpaced the tooling available to govern them. That lag is narrowing, starting with this deal.
The Cisco-Astrix Deal: What Happens Next
Cisco’s acquisition of Astrix Security, if it closes, will be remembered less as a $350 million bet on a single startup and more as the moment enterprise AI agent security became a named, funded, and institutionally owned product category. The companies that treat it as optional are taking the same position that late cloud security adopters took in 2017 — and they paid for it in breach costs and remediation spend.
Every enterprise software vendor now faces the same question: build agent monitoring capabilities before a major incident forces the issue, or wait until Cisco’s distribution makes the independent market irrelevant. The evidence from every prior security category consolidation points in one direction.
