LONDON — Biometric identity firm iProov on Wednesday launched the first commercial service built on the Human Approval and Presence Specification (HAPS), an open standard designed to make Human Approval for AI Agents cryptographically verifiable. The hosted HAPS approval service, announced on October 22, gives enterprises deploying autonomous agents a way to prove that the right human being genuinely approved a specific high-risk action before the agent was allowed to execute it.

The service launches with two first integrations: agentgateway, the open-source AI agent gateway developed by solo.io, and Cerbos, the authorization management platform. iProov chief technology officer Dominic Forrest is presenting the service and both integrations at AgntCon North America in San Jose on October 22 and 23, according to the company's announcement.

How the HAPS approval loop works

When a Cerbos policy flags an action as requiring human approval, execution is paused and iProov's hosted HAPS service is called. The service shows the named approver the exact action and its scope. If the approver agrees, they complete an iProov genuine human presence check using the company's Dynamic or Express liveness technology, the same technology relied upon by national governments and major organizations. iProov then issues a signed, single-use credential bound to that exact action, and only then is the agent's action released.

If any detail changes — the amount or destination of a payment, for example — the credential no longer matches and the action is blocked. The result is an auditable, non-repudiable certificate that iProov says no agent can produce on its own behalf. Together, the credential, the scope binding and the audit certificate form a practical recipe for Human Approval for AI Agents that regulators and auditors can inspect after the fact.

HAPS itself is proof-agnostic: the specification does not mandate biometrics or any particular presence check, and it is designed to supplement existing systems such as OAuth, OpenID Connect and WebAuthn rather than replace them. The open specification, a partial Rust reference implementation and test vectors are published at github.com/iProov/HAPS under the Apache-2.0 license, so any gateway, policy engine or relying party can build HAPS-compliant approval backed by iProov or by other forms of proof.

Importantly, HAPS does not demand a human sign-off for every routine task. Organizations set their own thresholds for which sensitive or high-stakes actions require confirmation, avoiding the approval fatigue that would undermine the benefits of agent autonomy.

Why verifiable approval is arriving now

The launch lands in the middle of a fast-moving week for agent governance. The Agent Control Standard (ACS), a vendor-agnostic open framework for runtime governance of AI agents, was announced during the AI Agent Security Summit in San Francisco, defining middleware hooks that let policy engines allow, deny or modify agent actions inline. In an October 2026 Microsoft Security blog, Microsoft highlighted the importance of policies that can require cryptographically provable human approval for a high-impact step — the exact capability iProov's service is built to deliver.

The gap HAPS addresses is subtle but large: an AI agent may hold legitimate permissions yet take an action its human never intended, through prompt injection, excessive goal-seeking or misused delegated credentials. Access does not equal intent. Cerbos chief product officer Alex Olivier said the partnership lets organizations require approval from the right human before a high-consequence action can proceed, giving them confidence to delegate a wider range of tasks to agents while retaining oversight and accountability. iProov founder and CEO Andrew Bud framed the same point in identity terms: requiring a genuine human face in the approval loop secures the process against agent impersonation, because no matter how smart agents become, they will never have faces.

Similar moves are appearing across the ecosystem. World and Vercel have introduced cryptographic human-in-the-loop approval for agent workflows, while banks have warned about fraud, scams and data-privacy risks as agents begin shopping and paying online. As retailers open checkout to agents and agentic payments expand, the question of who approved what — with proof — is becoming a procurement and compliance requirement rather than a nice-to-have. Verifiable Human Approval for AI Agents is what makes that requirement deployable inside real policy engines.

What to watch next

iProov is positioning HAPS as shared infrastructure rather than a proprietary product, inviting scrutiny, challenge and independent implementations. The specification was first published experimentally in September and presented at AGNTCon + MCPCon Europe in Amsterdam; this week's launch is the step from open specification to hosted service. Whether enterprises wire it into production agent gateways — and whether auditors accept the resulting certificates — will determine whether verifiable Human Approval for AI Agents becomes a standard layer of the agent stack.

Sources

Related coverage: Continue with the latest GenZ NewZ headlines, more ai news coverage, Gen Z Trusts AI Holiday Planning More Than Google, Sundae Bar AI Launches Sundae Personal AI Agent, and PHP Security Flaw Knocks Government Sites Offline.