Arkose Labs Launches Agent Trust Manager to Govern Undisclosed AI Agent Traffic
What Happened — Arkose Labs announced an expansion of its Agent Trust Manager platform, enabling enterprises to classify, monitor, and block AI‑driven “agentic” sessions that do not self‑identify. The new capabilities let organizations define step‑by‑step policies (allow, monitor, block) across the customer journey—sign‑up, login, checkout, money movement—and enforce them even when agents change signals or networks. Features include Agent Orchestration, Visibility & Classification (good vs. malicious agents), a stable Agent ID that follows an agent across IP changes, and Detection Auditing that records classification decisions for export. Global intelligence feeds compare a business’s agent mix with Arkose’s network‑wide data, and a forthcoming “Proof of Human” step will require explicit human confirmation for high‑risk actions.
Why It Matters for Trust & Control Assurance — The control‑objective of AI system oversight—continuous classification, policy enforcement, and audit‑ready evidence—directly supports emerging AI governance frameworks such as the NIST AI RMF.
Who Is Affected — SaaS security providers, financial services platforms, e‑commerce sites, and any organization that processes high‑value transactions through web or mobile channels.
Recommended Actions
- Map your existing bot‑management controls to the new agent classification and policy layers.
- Deploy detection‑auditing logs to create defensible evidence for AI‑governance audits.
- Integrate the Agent ID concept into your identity‑management workflow to maintain continuity across network changes.
Technical Notes
- Agent classification tiers: self‑disclosing good, non‑disclosing good, malicious.
- Agent ID is a stable identifier independent of IP address.
- Detection Auditing provides an exportable audit trail of signal updates and classification outcomes.
- No disclosed CVEs; the release is a product‑feature update.
Source: Help Net Security