The Standard

AI Human Proof:
The Governance
Standard

To be AI Human Proof is to have verified that every AI system in operation meets the standard of checks and balances that professional responsibility for the AI era requires. This page is the canonical definition of that standard.

The Definition

AI Human Proof (adjective): Describes an AI system, and the organisation operating it, that has implemented the full set of governance controls required to deploy AI responsibly — including pre-deployment risk assessment, operational accountability, continuous performance measurement, and incident response capability.

The term is constructed deliberately. "Human Proof" does not mean proof against humans — it means proof by humans: the verification, by people, that an AI system's behaviour is understood, monitored, and controllable. In an era in which AI systems increasingly make decisions that humans once made, human oversight is the essential check on automated power.

An organisation that is AI Human Proof does not merely have an AI ethics policy. It has the operational controls, documented processes, and named accountability that allow a reasonable external observer to verify that its AI deployments meet the standard of professional responsibility. The distinction between documented intent and operational capability is the defining difference between governance and theatre.

"The question for every organisation deploying AI is not 'do we have an AI ethics policy?' It is 'can we demonstrate, to a sceptic, that our AI systems are under genuine human control?'"

Why This Standard Exists

Five reasons the AI Human Proof standard is necessary — not aspirational, necessary.

  1. AI systems make consequential decisions

    Modern AI is not a tool that assists human decision-making — it is increasingly the primary decision-maker in hiring, lending, healthcare triage, content moderation, and criminal justice. Systems that make consequential decisions require checks. This is not a new idea. It is the foundational principle of every regulated industry.

  2. The liability is already real

    The EU AI Act, US Executive Order on AI, UK AI Regulation Policy, and a growing body of litigation establish that organisations deploying AI carry legal and ethical responsibility for what those systems do. "We didn't know the model was biased" is not a defence. "We didn't build the oversight mechanisms" is evidence of negligence.

  3. Most governance is performative

    Published AI ethics principles, responsible AI teams, and governance frameworks are widespread. Operational governance — the specific, auditable controls that actually constrain what AI systems do — is rare. The gap between what organisations say about AI governance and what they have built is the space where harm occurs.

  4. The industry has no independent standard

    NIST AI RMF, ISO 42001, EU AI Act, and dozens of voluntary frameworks each address part of the governance problem. None of them provides a single, clear, deployment-level standard that an organisation can use to verify that its AI deployment meets the bar of professional responsibility. AI Human Proof is that standard.

  5. Comprehensiveness is the only honest approach

    Partial governance is not governance. An organisation that has bias testing but not incident response, or human review but not audit logging, or a published ethics policy but no operational controls, has not implemented AI governance — it has documented an intent to govern. The AI Human Proof standard specifies the complete set of requirements because partial sets produce partial protection.

What Meeting the Standard Requires

Four non-negotiable capabilities. All of them, not some of them.

  1. Pre-deployment verification

    Every AI system receives a structured risk assessment before deployment. Use case, affected populations, failure modes, and data quality are documented and evaluated by someone other than the development team. Systems with unacceptable risk profiles do not deploy.

  2. Operational accountability

    A named individual — not a committee, not a policy document — holds accountability for each deployed AI system. This person has the authority to suspend or terminate the deployment. Their name is on the System Card.

  3. Continuous measurement

    Bias evaluations, accuracy assessments, and fairness metrics run on a defined cadence — not only at launch. Alert thresholds are set before deployment. When thresholds are exceeded, a defined response process activates.

  4. Incident response capability

    When the system produces harmful, inaccurate, or policy-violating outputs, the organisation has a documented, tested process for detecting, triaging, responding, and learning. Detection is active, not reactive. Response timelines are defined.

Questions About the Standard

Is "AI Human Proof" a certification?
AI Human Proof is a governance standard — a defined set of requirements for responsible AI deployment. It is not currently a third-party certification programme, though the standard is designed to be auditable. Organisations can self-assess against the standard and use it as a framework for external review.
Does the standard apply to every AI system?
The full standard applies to AI systems that make or influence consequential decisions affecting people. A spam filter and a hiring algorithm both use AI; they do not carry the same governance obligations. The standard includes a risk tiering methodology for determining which requirements apply to which systems.
How does AI Human Proof relate to NIST AI RMF and the EU AI Act?
AI Human Proof is operationally compatible with both NIST AI RMF and the EU AI Act. Where NIST provides a reference architecture and the EU AI Act provides a legal framework, AI Human Proof provides the deployment-level specification — the specific controls, documentation, and processes that satisfy both. The Hub documents those relationships in detail.
Who is AI Human Proof for?
Any organisation deploying AI systems that affect people: enterprises, public sector organisations, AI vendors whose tools are used in consequential decisions, and the professional advisors (legal, technical, operational) who support them. The standard is written for people who build and operate AI systems, not for those who evaluate them from outside.
How is the content on AI Human Proof produced?
Every piece of content is built from primary sources — regulatory texts, peer-reviewed research, official model documentation, and first-hand case study evidence. The Hub's value is in re-authoring that material into a structured, accessible format that preserves the intellectual substance while making it operationally useful. Sources are cited on every article.
What does "Human Proof" actually mean?
It means proof by humans, not proof against them. The term describes the verification, by people, that an AI system's behaviour is understood, monitored, and controllable. As AI systems take on decisions humans once made, human oversight becomes the essential check on automated power — and "AI Human Proof" names the state of having built that check.
Can we self-assess against the standard?
Yes — self-assessment is the intended first use. The standard is written to be applied by the people who operate the systems, and organisations are encouraged to reference it in their own governance documentation. Where an assessment needs to withstand external scrutiny, self-assessment is the preparation for that review rather than a substitute for it.

How to implement the standard

The Hub contains everything you need to build AI Human Proof governance in your organisation.

Frameworks

Governance Architectures

The methodologies and regulatory standards that define AI governance practice.

→Tools

AI Platform Audits

What commercial AI platforms provide, and what your organisation must build.

→Guides

How-To Implementation

Step-by-step guidance for building the operational controls the standard requires.

→