Arkose Labs Expands Agent Trust Manager to Detect and Govern Non-Disclosing AI Agent Traffic
Arkose Labs has expanded its Agent Trust Manager product to help businesses classify and control automated AI agent sessions across customer journeys such as sign-up, login, checkout, and money movement. The update is driven by findings from Arkose's State of Agentic AI Trust Q3 report showing that personal AI agents self-identify less than 3% of the time, making disclosure-based detection ineffective. New features include per-step agent orchestration policies (allow/monitor/block), agent visibility and classification (self-disclosing good, non-disclosing good, malicious), a persistent Agent ID that follows the agent across network and build changes, Detection Auditing for signal history, global agentic traffic intelligence, and an upcoming Proof of Human step for sensitive actions like money movement. This is a product announcement rather than a disclosed incident or vulnerability.
Detalhes técnicos
The platform integrates with Arkose Titan to enforce policies in real time and exposes querying via Ask Titan and an AI Assist (MCP) integration for plain-language traffic analysis within Command Center. Detection Auditing provides an exportable audit trail showing when detection signals were updated and how sessions were classified at each point in time, intended to support compliance and investigation workflows.
Linha do tempo
-
Arkose Labs Expands Agent Trust Manager to Detect and Govern Non-Disclosing AI Agent Traffic
ler a análise desta fonte ocultar
Arkose Labs has expanded its Agent Trust Manager product to help businesses classify and control automated AI agent sessions across customer journeys such as sign-up, login, checkout, and money movement. The update is driven by findings from Arkose's State of Agentic AI Trust Q3 report showing that personal AI agents self-identify less than 3% of the time, making disclosure-based detection ineffective. New features include per-step agent orchestration policies (allow/monitor/block), agent visibility and classification (self-disclosing good, non-disclosing good, malicious), a persistent Agent ID that follows the agent across network and build changes, Detection Auditing for signal history, global agentic traffic intelligence, and an upcoming Proof of Human step for sensitive actions like money movement. This is a product announcement rather than a disclosed incident or vulnerability.
Technical Details
The platform integrates with Arkose Titan to enforce policies in real time and exposes querying via Ask Titan and an AI Assist (MCP) integration for plain-language traffic analysis within Command Center. Detection Auditing provides an exportable audit trail showing when detection signals were updated and how sessions were classified at each point in time, intended to support compliance and investigation workflows.