Trend Micro Identifies Four Agentic AI Controls Enterprises Are Missing: Inventory, Least-Agency, Supply Chain, and Communication Monitoring

What happened

Why it matters

• ·Agents operating without an inventory create an immediate audit and compliance gap: organizations cannot demonstrate control over systems they cannot enumerate, which becomes directly problematic under frameworks such as the EU AI Act Implementation Timeline and emerging sector-specific agentic rules signaled by regulators like the Bank of England.
• ·The supply-chain framing for AI tools and extensions expands the vendor risk surface significantly, because a compromised or malicious plugin can inherit an agent's permissions and act at scale across enterprise systems, as illustrated by incidents like the Meta Sev-1 Agent Incident where authorization failures enabled unexpected lateral action.
• ·Communication-flow monitoring between agents is an operationally new requirement that existing security information and event management programs were not designed to handle, meaning compliance teams may face a gap between what they attest to auditors and what their tooling can actually detect.

Governance controls affected

What to do now

• ☐Conduct an agent inventory exercise to enumerate every deployed agentic system, including those embedded in third-party platforms, documenting owner, task scope, and permission set for each.
• ☐Review agent permission configurations against a least-agency standard, confirming that each agent can access only the data sources, APIs, and actions its specific task requires and that no standing broad permissions remain.
• ☐Classify all AI tools, plugins, and extensions connected to agentic systems as supply-chain risk items and route them through your existing third-party AI risk assessment process.
• ☐Extend your monitoring architecture to capture inter-agent communication flows, not only endpoint outputs, and establish behavioral baseline thresholds that trigger review when agent activity deviates.
• ☐Define and document which autonomous agent actions require a human approval gate before execution, with particular attention to actions that modify data, initiate financial transactions, or are difficult to reverse.

What to watch next

Regulatory pressure on agentic AI controls is accelerating from multiple directions. The DHS and CISA guidance on AI agents in critical infrastructure is likely to inform sector-specific rulemaking, and the Bank of England's signals on bespoke agentic AI rules suggest that financial services firms will face mandatory autonomy controls within the next regulatory cycle. Compliance teams should also watch for whether the NIST AI RMF Playbook is updated to incorporate agentic-specific guidance, and monitor whether the Illinois AI Safety Measures Act audit requirements, which took effect alongside China's agent rules, produce early enforcement patterns that set a de facto standard for agent documentation and inventory practices.