iProov Launches Human Approval Service for AI Agents
The service adds a genuine human presence check before enterprise AI agents can carry out sensitive actions.
Updated on Nov. 2, 2026 in Robotics
iProov has launched a human approval service for enterprise AI agent actions, integrating with agentgateway from solo.io and with Cerbos, its first policy-engine partner. When a Cerbos policy requires sign-off, the agent's action is held until a named approver completes an iProov human presence check.
Why it matters
Some actions must remain subject to human control, approval and accountability as AI agents take on more tasks. The service addresses that need by binding approval to a specific, verifiable person and action.
The HAPS engine uses iProov's Dynamic and Express liveness technologies and issues an auditable, non-repudiable certificate that can be logged. Iproov issues a signed, single-use credential bound to the exact action, and the credential fails if any detail, such as a payment amount or destination, changes.
The players
iProov
A biometric identity verification company known for its liveness detection technology, which underpins the new approval service.
solo.io
A software company that develops agentgateway, an open-source AI agent gateway the service integrates with.
Cerbos
An authorization management platform that determines when human approval is required and triggers the HAPS service.
The details
The service implements the Human Approval and Presence Specification (HAPS), released in September. When a Cerbos policy requires sign-off, the hosted HAPS service shows the named approver the exact action and its scope before they complete the check.
Timeline
September (prior year): HAPS specification released.
November 2: iProov's human approval service launch reported.
The Tech Race
The launch operationalizes the Human Approval and Presence Specification (HAPS), released in September, as a commercial product. It positions iProov's liveness technology as a control layer for autonomous AI agents across enterprise software stacks.
Enterprise teams deploying AI agents can require a named human to approve sensitive actions before they execute. Each approval is logged as an auditable certificate, which can support accountability reviews.
The takeaway
Keeping a verifiable human in the loop for high-stakes AI actions is becoming a practical requirement rather than an optional safeguard. Teams building agent systems should plan approval checkpoints before granting agents access to payments or other sensitive operations.
Further reading
For broader context on autonomous systems and human oversight, see Robotics.
