Organisations are faced with a new security quandary as they place more onus on their AI agents: how to be sure these agents are in fact doing what they should. While shielding them from cyberattacks is part of the problem, it is just as vital to comprehend their decisions, keep tabs on their actions and spot any unanticipated behaviour. CrowdStrike intends to deal with both aspects through its forthcoming purchase of Fiddler AI, an outfit specialising in AI observability and explainability.
The transaction was made public on November 3 at Fal.Con Europe 2026 in Barcelona. Under the terms of the deal, Fiddler’s technology will be incorporated into the Falcon platform as Falcon Observe, with general availability set for the fourth quarter of FY27. The purpose is to link security with an ability to see what an agent is up to, so enterprises can determine if an agent is not only secure but also trustworthy for the tasks at hand.
What Fiddler AI Brings to Enterprise AI Agent Security
Headquartered in Palo Alto, California since its 2018 founding, Fiddler AI has built out technology to show businesses what their AI systems are doing and the reasoning behind particular decisions, as well as if they are staying within bounds.
That extends over the full life of an agent. In the pre-deployment stage, an enterprise can judge accuracy, policy compliance and task completion against a range of models and prompts. When the agents are running, traceability tools follow the tool calls and decision making down the hierarchy of agents for teams to review in the event of a failure.
Fiddler also keeps an eye on production behaviour to catch drift or hallucinations and puts in place auditable records for first and third party as well as coding agents. Then there are the Centor evaluator models. They measure safety, faithfulness and sensitive information at a much lower cost than large language models (LLMs) and do so from within the customer’s cloud or VPC.
How Falcon Guardian and Falcon Observe Complement Each Other
With this acquisition two things come together. Fiddler gives a view of the agent’s instructions and outcomes; CrowdStrike adds the security context of the identities, endpoints and cloud resources involved.
Falcon Guardian, which was announced with the acquisition, is aimed at defending agents from misuse and attacks while they are running and features an AI Kill Switch to put an end to any unauthorised agents on the spot. Falcon Observe will be there to confirm the agent is acting as it should. Put together, the idea is to let an enterprise have a say in three matters: which agents are fit for deployment, what they are allowed to do and when they can be given more responsibility.
CrowdStrike-Fiddler AI Acquisition: Deal Value and Expected Closing
No financial details have been put forward by CrowdStrike regarding the Fiddler AI buyout. But according to SecurityWeek and The Wall Street Journal, the company’s acquisitions of Fiddler and Dreadnode come to some $495 million in total, with the lion’s share of $400 million going to Fiddler and $95 million to Dreadnode. Most of that will be in cash, with stock subject to vesting making up the rest. The close is due in the fourth quarter of FY27, ending 31 January 2027, provided customary conditions are met.
Why AI Agent Behaviour Monitoring Matters for Enterprises
This is indicative of a change in the way enterprise AI security is handled. With ever more autonomous agents in the field, protection from outside threats is not enough; there has to be oversight of decisions and accountability.
Fiddler’s co-founder notes the firm was set up eight years back for precisely that reason, to give business a handle on the AI they rely on. Its work has since branched into model monitoring, runtime guardrails and agentic observability among other things. For current Fiddler clients, the firm says the combined product will still be about governance, cost and quality.
For the business at large the message is plain: the scaling of AI agents will be as much a matter of being able to govern and secure their conduct as of what they are capable of.
