HYPR and Hitachi Unveil Ways to Tie AI Agent Actions to a Verified Human

HYPR has launched AgentPass, a product that binds each AI agent in an organization to a named human supervisor, while Hitachi has developed a prototype that uses biometrics to prove a person delegated a task to an AI agent. Both aim to show, at the moment an agent acts, which person authorized it.

HYPR AgentPass, available now, assigns every agent to a supervisor who signs in through the organization’s identity provider. All agent connections run through a single gateway where agents start with no access, see only the tools allowed by their supervisor’s role, and never hold the credentials for those tools. The gateway keeps the credentials and issues short-lived session tokens that work for a single tool. Because access follows the supervisor’s role, a role change updates the agent’s access, and offboarding the supervisor cuts the agent off on its next request.

High-risk actions and new agent connections are routed to the supervisor’s mobile device, separate from the machine the agent runs on, for approval with a FIDO2 passkey, so the person signs the exact action requested. Authorized users can disable an agent from the console or the HYPR mobile app, and its next request is refused everywhere. AgentPass also discovers AI tools and coding agents such as Claude Code, Codex and Cursor on enrolled macOS and Windows workstations, and checks every request against policies that administrators can write in plain English and test in an “Observe” mode before enforcing them.

“Security teams cannot protect what they cannot govern,” said Bojan Simic, CEO and co-founder of HYPR. “AgentPass brings the same verifiable identity assurance we provide for human workforces directly to agents, ensuring that every agent action is authorized, bound by policy, and tied to an accountable individual.”

Hitachi’s prototype, called Proof of Human Delegation, attaches a digital signature to a letter of delegation that sets out what an AI agent may do on a person’s behalf, such as a transaction amount or payee. The key used to sign that letter is generated from the person’s biometric data each time a signature is made. Verifying the signature with the matching public key confirms that the person delegated the task, that the letter has not been altered, and who delegated what and to what extent. A business receiving a request from the agent can compare it against the letter before accepting it.

The technology is built on Hitachi’s Public Biometric Infrastructure (PBI), which combines biometric authentication with public key cryptography. At enrollment, PBI creates a key pair from biometric data, stores the public key in the cloud and discards the private key; the private key is then recreated from the person’s biometrics for each authentication and discarded after use. Hitachi designed Proof of Human Delegation to make identity authentication and signing inseparable: when an AI agent strays from a person’s intent, or an attacker misuses a stored private key, credentials or a signing function, a delegation letter the person never approved can otherwise still carry a valid signature.

The prototype was built for a procurement scenario and covers generating and registering PBI public keys, signing delegation letters with identity verification, and verifying signed letters. Hitachi plans to evaluate it for financial transactions and payments, industry, mobility and energy, and to define signing and verification APIs, letter specifications and requirements for connecting to existing identity verification and access control systems.

Sources: GlobeNewswire, Hitachi, IT Leaders

–

By the ID Tech Editorial Team