iProov, the world’s leading provider of biometric identity verification solutions, published the Human Approval and Presence Specification (HAPS), an experimental procedural specification for managing AI Agents. HAPS is designed to help organizations verify that a human has approved a specific action before an AI agent is allowed to carry it out.
Published on GitHub under the Apache-2.0 license, HAPS consists of a set of rules developed in direct response to the growing need for stronger governance and human oversight as AI agents become more capable of acting autonomously on behalf of their human approvers. The specification includes a partial Rust reference implementation and test vectors; iProov invites industry experts to review and critique the specification, provide feedback, and explore independent implementations.
While containment failures like the July 2026 OpenAI Hugging Face breach draw the headlines, HAPS addresses a quieter problem: the agent that stays within its permissions yet takes unintended actions, whether through prompt injection, excessive goal-seeking, or misused delegated credentials. The danger is that the receiving system has no way to determine the legitimacy of the request.
This highlights a key gap in AI governance: AI agents can access tools and services, but that does not prove a human intended or approved the actions they take. As a prototype framework, HAPS is designed to let organizations and relying parties enforce verifiable human approval before critical actions can proceed. “AI agents are moving rapidly from answering questions to taking actions on our behalf. As their autonomy and capabilities grow, governance must keep pace,” said Andrew Bud, founder and CEO of iProov. “We need to distinguish between an agent having permission to act and a human actually approving the specific action it is about to take.
HAPS is a specification that makes that human approval verifiable. We’re publishing it openly at this experimental stage because the industry needs to solve this challenge together. We’re inviting the community to scrutinize it, challenge it, and build on it so together we can establish strong, practical safeguards for an agentic AI world.”
HAPS is not intended to require human approval for every action an AI agent takes, as doing so would undermine the benefits of autonomy and risk creating approval fatigue. Instead, organizations decide which actions are sensitive and/or critical enough to require additional human approval and what level of evidence is appropriate.
When additional approval is required, the principle is simple: pause the action, show the human exactly what the agent wants to do, obtain proof of genuine human presence and approval, and verify that evidence before allowing the action to proceed. HAPS securely links the human’s approval to the specific action, allowing the organization to check that the approval matches what the agent is actually asking to do.
HAPS is proof-agnostic and does not prescribe how human presence must be proven. iProov has developed its own internal implementation using biometric liveness as one example of how the specification could be applied in practice as one example of an agent-resistant proof.
