Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

Most organizations now have an AI policy on paper. Far fewer can tell whether employees follow it. According to ISACA, 25% of organizations have no AI policy at all, and only 38% of those with one describe it as formal and thorough. Meanwhile, employees keep pasting, uploading, and delegating work to AI tools, both generative and agentic. Cyberhaven Labs found that 39.7% of employee interactions with AI tools involve sensitive data.

As organizations deploy increasingly autonomous and interconnected agents, the potential agentic blast radius grows. A single compromised or misaligned agent can trigger a chain of actions spanning credentials, applications, files, and enterprise systems that create consequences extending far beyond its original task. Traditional security approaches weren't designed to understand these behaviors, much less intervene at the speed agents operate. Security teams need more than alerts and investigations.

I am delighted to announce CrowdStrike’s intent to acquire Fiddler at Fal.Con EU 2026. Fiddler, a leader in AI observability, will help ensure enterprise AI agents act as intended with pre-deployment evaluation and continuous behavioral monitoring as part of the CrowdStrike Falcon platform. Every business deploying AI agents asks the same questions: What are our agents doing? Are they staying on task? Are they acting in a way that could put the business at risk?

When Claude Cowork executes code locally, it runs inside an isolated virtual machine on the user's computer. On the host, that VM runs as a single opaque process: The agent's activity inside the VM, including the processes it spawns, the packages it installs, and the network connections it makes, is not visible to the host's endpoint security sensor. Because host folders can be mounted into the VM, sensitive data can sit within the agent's reach.