TrendAI Extends Claude Audit Coverage: Security Teams Can Now Investigate Every AI Agent Action
TrendAI Vision One now ingests Claude Code and Cowork session records, giving security teams full audit trails of AI agent prompts, responses, and tool calls.
The headline
TrendAI, the AI security business unit of Trend Micro, announced on September 30 that it has expanded its Claude Compliance API integration in TrendAI Vision One. The integration now covers Claude Code and Cowork session transcripts from Claude Enterprise — including prompts, responses, and tool calls, each tied to a verified user. It builds on the integration TrendAI announced with Anthropic in June 2026.
The visibility gap this closes
AI agents now write code, call tools, and reach into business systems on behalf of employees. Security teams usually know these agents are running, but not what they were asked to do, what they did, or whether it followed policy. That gap is exactly where enterprise AI risk lives: an agent with production credentials and no audit trail is a liability no compliance framework accepts.
With this expansion, security teams can bring AI session records into Vision One and investigate them in context — correlating AI activity with endpoint, identity, cloud, and network signals, and building custom detections for the agent behavior that matters. In practice, that means teams can flag agent tool calls outside expected boundaries, tie AI session activity to specific users during investigations, check whether agent workflows follow required processes, and keep an auditable record for compliance review.
"Coding and knowledge-work agents are now doing real work inside the business, and security teams should be able to see and investigate that work the same way they do everything else," said Rachel Jin, Chief Platform and Business Officer and Head of TrendAI.
Agent security is becoming infrastructure
This is the second major agent-security move this week: TrendAI also extended NVIDIA's Agent Safety Platform with threat intelligence and hardware-level detection on BlueField DPUs. The pattern is clear — agent security is splitting into layers. The infrastructure layer (NVIDIA's OpenShell, hardware watchdogs) enforces policy where the agent cannot bypass it; the intelligence layer (TrendAI, compliance APIs) provides the visibility and audit trail that tells you what actually happened.
Enterprises moving agent fleets from pilot to production are discovering that security is often the deciding factor. No single control makes an agent secure; it has to be built into every layer the agent depends on — model, orchestration harness, and the tools and data the agent can reach.
What this means for your business
If your teams use Claude Code, Cowork, or any coding agents against company systems, you now have a credible answer to the auditor's first question: show me what the agent did. Before this class of integration, most companies had activity logs at best and vibes at worst.
The operational lesson: deploy AI agents with the same observability you demand from human employees with production access — identity-attributed actions, reviewable logs, and anomaly detection. Agents that act without audit trails are technical debt that compounds into compliance risk.
What to do next
Inventory every AI agent currently touching your code, data, or customer systems, and check whether its actions are logged against a verified user identity. If any agent operates without an audit trail, that is your highest-priority AI security gap — close it before you scale from pilots to fleets. For regulated industries, make session-level AI auditability a procurement requirement in your next vendor review.