Cisco to buy Galileo for AI observability
Cisco announced plans to acquire Galileo, an AI observability startup, to fold its capabilities into Splunk and strengthen visibility and protection for AI agents and models. The move signals observability is becoming a core layer of the AI stack as enterprises demand runtime monitoring, audits and controls for production models. (networkworld.com)
Cisco is buying a startup that watches artificial intelligence the way a flight recorder watches a plane: not just whether it ran, but what it saw, what it decided, and where it went wrong. The startup is Galileo, and Cisco said on April 9, 2026 that it plans to bring Galileo into Splunk, the software company Cisco bought for about $28 billion in 2024. (cisco.com 1) (cisco.com 2) That tells you where the market is moving. Companies no longer just want artificial intelligence models that answer questions in a demo; they want systems they can inspect in production after a bad answer, a blocked request, or an automated action that should never have happened. (networkworld.com) (galileo.ai) Galileo was built for that job. Its platform says it helps teams evaluate, monitor, and protect generative artificial intelligence applications and agents, which are software systems that can take multiple steps on their own instead of replying once and stopping. (galileo.ai) (siliconangle.com) Splunk’s old strength was machine data: logs, alerts, and traces from servers, apps, and networks. Cisco’s bet is that artificial intelligence systems now need the same treatment, except the “machine” is a model that can hallucinate, leak data, or call the wrong tool while sounding confident. (cisco.com) (networkworld.com) Cisco said Galileo will extend Splunk from traditional observability into model and agent monitoring. In plain English, that means a company using an internal sales agent or coding agent could track which prompt triggered a risky answer, which retrieval step pulled bad context, and which action chain touched a sensitive system. (cisco.com) (crn.com) The timing was not random. In March 2026, Galileo announced Agent Control, an open-source control plane for governing agent behavior in real time, and Cisco AI Defense showed up as a launch partner, so the two companies were already working on runtime guardrails before the acquisition announcement. (galileo.ai 1) (galileo.ai 2) That matters because the hard part of enterprise artificial intelligence is shifting from model training to runtime control. A chatbot that drafts text is one risk; an agent that can search internal files, open tickets, or trigger workflows creates a bigger one, because one bad decision can turn into an actual business action. (galileo.ai) (cisco.com) Galileo has been pitching itself as more than a dashboard company. Its site says it turns offline evaluations into production guardrails, which means tests done before launch can become live rules that catch failures after launch instead of sitting in a slide deck. (galileo.ai) Cisco is also stitching this into a larger strategy that has been visible since the Splunk deal closed. Cisco has been arguing that networking, security, and observability data belong in one place, and artificial intelligence systems give that argument a new use case because the same agent can touch all three layers in one session. (cisco.com 1) (cisco.com 2) You can read this acquisition as Cisco saying artificial intelligence observability is no longer a side tool for experimental teams. Once companies start putting agents into customer support, software operations, and internal workflows, the audit trail becomes part of the core stack, and Cisco wants Splunk to be where that trail lives. (networkworld.com) (cisco.com)
