AI governance moves fast. We track it daily.

Krebs on Security published a deep-dive on the attack surface created by autonomous AI agents, documenting a "lethal trifecta" where AI assistants combine private data access, untrusted content exposure, and external communication capabilities into a single exploitable surface. The piece catalogs real-world incidents including 600+ compromised FortiGate appliances across 55 countries, a supply chain attack on the Cline coding assistant via ClawHub, and the Moltbook platform hitting 1.5 million agent registrations in a single week with emergent autonomous behaviors.

On the data sovereignty front, the FBI confirmed it actively purchases commercial location data to track American citizens without warrants. This reinforces the core governance message: data flows through pathways organizations don't anticipate or control. The warrant bypass via commercial data brokers is a clean analogy for how employee AI tool usage creates unintended data flows. If you don't control where your data goes, someone else decides who sees it.

Today's two articles represent different layers of the same story: the Krebs piece maps the technical attack surface (agents with too much access) while the FBI piece illustrates the downstream consequence. Together they give operational leaders a compelling one-two for internal conversations about AI governance: your AI tools are creating data flows you can't see, and once that data is out, even law enforcement can access it without a warrant.