OISTE Foundation, home
  1. oiste.org
  2. AI and Identity
  3. HUMAN-AI-T

AI and Identity

HUMAN-AI-T

A governance framework for artificial intelligence that stays under human control. Its principles are political; its enforcement is cryptographic: an AI system that cannot be identified cannot be governed, and one whose authority cannot be revoked cannot be stopped.

Human-Centered Artificial Intelligence and TechnologyLaunched 2025

The problem

The internet was built without identity. AI cannot be.

Language models produce text, images and voices with no anchor to a real-world source. Without a verifiable identity layer there is no reliable way to tell a person from a synthetic actor, an authorised agent from a hijacked one, or an authentic document from a fabricated one.

That is a missing layer in the architecture, and it grows more serious as AI moves from generating information to taking actions: moving money, signing contracts, calling APIs, operating machines.

HUMAN-AI-T proposes to supply that layer from the root of trust OISTE already holds, and to make identity, revocation and provenance the obligations any binding instrument on AI will need.

Principles

Seven principles

As set out in the HUMAN-AI-T manifesto presented at Davos in January 2026.

  1. Human control

    Humans remain in control of AI systems, with meaningful oversight and decision authority.

  2. Support, not replacement

    AI supports human decision-making rather than replacing it, particularly in high-impact domains.

  3. Values by design

    Human values are deliberately built into algorithms, not assumed to emerge from data.

  4. Harm prevention

    Prevention of harm to health, security, human rights and the environment is embedded by design.

  5. Fairness

    Fairness and non-discrimination are enforced, so that AI treats all people equally.

  6. Transparency

    Transparency, explainability, auditability and traceability are mandatory requirements.

  7. Rights protection

    Privacy, civil rights and data sovereignty are protected as fundamental human rights.

Enforcement

Three obligations the root of trust can enforce

Principles need a mechanism. HUMAN-AI-T places it in identity infrastructure that already exists and is already audited.

Identity

Know Your Agent

The counterpart to Know Your Customer. Before an autonomous agent transacts, its issuer, operator and permitted scope can be verified by the party on the other side.

Revocation

Authority that can be withdrawn

An agent whose credential is revoked is refused at every layer it touches, without anyone needing access to its weights or code.

Provenance

Signed at the source

Content and actions are signed at the point of creation and chained to the root, so authenticity becomes a property of the file rather than a judgement about it.

How the credentials work

Human dignity as the test

Judge these systems by one question: whether they advance or diminish human dignity.

Carried into OISTE’s UN advocacy from The TransHuman Code

Milestones

  1. 2025

    Launch

    OISTE, WISeKey, SEALSQ and the UN Alliance of Civilizations launch HUMAN-AI-T.

  2. Apr 2025

    Presented at the United Nations

    Third UNAOC dialogue on AI for One Humanity, Geneva.

  3. Dec 2025

    World Human Rights Concert

    Presented at the Palais des Nations with ONUART.

  4. Jan 2026

    Manifesto

    Seven principles presented at Davos.

  5. Sep 2026

    Post-quantum identity for AI

    The root of trust issues credentials to AI models, agents and operators.