Post-quantum root of trust extended to AI models and agents

The architecture gives AI models, agents and their operators identities issued under the OISTE/WISeKey root of trust, using the NIST-standardised post-quantum algorithms ML-DSA and ML-KEM, with keys held in hardware security modules, trusted platform modules and secure elements.

It makes seven things verifiable: which model a system is dealing with; that the model is unchanged since release; that an agent is who it claims to be; that a named person authorised it; that two agents can trust each other; which agent signed a high-impact action; and the audit trail of all of the above.

Why it matters

AI is moving from generating information to acting on its own. Before an agent reaches infrastructure, talks to another AI or executes a transaction, the counterparty needs to know cryptographically who the agent is, who authorised it and whether it has been modified.

The work gives the HUMAN-AI-T framework its enforcement layer: identity, revocation and provenance anchored in a neutral, audited root.

Background

The Quantum Root Key was generated on 7 May 2025. It runs beside the foundation’s classical RSA and ECC roots so that authorities can migrate without a cut-over day.

Full release on GlobeNewswire