On October 22, 2026, iProov announced a hosted human approval service via Business Wire, built to implement the Human Approval and Presence Specification (HAPS). It is the first shipping implementation of a promise this beat has tracked only as a claim: cryptographically verifiable proof that a specific human authorized an AI agent’s action. We documented the gap in Unpriced Reviewer – enterprise harnesses have been promising exactly this while the proof stayed unpublished and the reviewer stayed unpriced.
The mechanism works like a gate on high-consequence steps. When an agent tries to execute a sensitive action, a policy engine – here, Cerbos – holds it. iProov’s hosted HAPS Approval Engine then shows the named approver the exact action and its scope. If that person approves, they complete an iProov Genuine Human Presence face check. iProov issues a signed, single-use credential bound to that exact action, and only then does the action run. Change any detail afterward – a payment amount, a destination – and the credential no longer matches, and the action is blocked. For security teams running human-in-the-loop workflows, the output is an auditable, non-repudiable record of who agreed to what.
Two integrations ship first. agentgateway, the open-source gateway built by solo.io, is the gateway side. Cerbos, an authorization platform that governs what identities and AI agents may do at runtime, is the first policy-engine partner. Cerbos decides when human approval is required, then calls the HAPS service.
Andrew Bud, founder and CEO of iProov, said in the release: “Our new HAPS service provides cryptographically verifiable proof that the real, right human being, confirmed by iProov Genuine Human Presence, has approved a specific action. Cerbos determines when such human approval is required and calls the HAPS service. Together, we enable organizations to embrace AI agent automation without compromising human control. By requiring a genuine human face as part of the approval process, we secure the process against agent impersonation. Because no matter how smart they become, agents will never have faces.”
Alex Olivier, Chief Product Officer and co-founder of Cerbos, said: “Cerbos controls and secures what decisions AI Agents can make in an organization, and under what conditions. Our partnership with iProov introduces a new level of control, enabling organizations to require approval from the right human before a high-consequence action can proceed.”
Buyers should weigh what sits underneath. The HAPS specification is published on GitHub under an Apache-2.0 license, but the repository’s own status line calls it an “experimental draft specification and partial reference implementation” that is “not production-ready.”
The October 22 announcement also leaves practical questions open: no pricing, no general availability date, and no customer deployment list appears in the release. Security teams still have to decide whether a face check stops an agent that can already act, or whether it adds a verification layer to a decision heading for execution anyway. The release arrived days after Microsoft’s October 19 Security blog argued that policies must “require cryptographically provable human approval for a high-impact step,” and iProov presented the integration at AgntCon North America in San Jose on October 22-23. Four days after the fact, the promise this family has been tracking has a working artifact to test – experimental, unpriced, and real.