qodiqa
SpecificationRun a simulation
OECD AI Principles (identification only)

Frameworks · Standards and guidance

qodiqa and the OECD AI Principles

qodiqa checks each AI request before the call and keeps tamper-evident, offline-verifiable evidence. This page shows where that evidence supports the OECD AI Principles, and where it does not.

  • Guidance
  • OECD
  • Adopted 2019, revised May 2024.
  • Verified 2026-11-01
Read the official textBack to all frameworks

Overview

What it is

OECD AI Principles (identification only)
Issuer
OECD
Type
Guidance
Status
Adopted May 2019; revised May 2024 (OECD/LEGAL/0449)

OECD Recommendation of the Council on Artificial Intelligence (OECD/LEGAL/0449). The OECD AI Principles are the first intergovernmental standard on AI. Section 1 sets five value-based principles for trustworthy AI, including transparency and explainability, robustness, security and safety, and accountability. They are not binding law, but many national rules draw on them.

Who it applies to. All AI actors, according to their roles. Governments that adhere to the Recommendation are asked to promote the principles, and the Recommendation addresses organisations only indirectly.

Where it meets qodiqa

The parts of the OECD AI Principles where the evidence matters.

Traceability

Request hash, audit entry and transparency leaf tie a decision to one request.

Accountability

Records a third party can verify, offline.

Human oversight

Revocation and a kill switch let a system be stopped and recorded.

Transparency

A disclosure block makes AI use visible on every executed response.

Robustness and security

Fail-closed checks and signed, single-use artifacts.

Risk management

Counts and reasons support a systematic, ongoing risk management approach.

Mapping

What qodiqa provides, and what stays with you.

Requirement or controlWhat qodiqa providesWhat stays with you
1.2(b) Mechanisms and safeguards, such as capacity for human agency and oversight, including for misuse.Revocation, a kill switch and human approval for irreversible tool actions are safeguards at the point of execution.The wider design of human agency in your product.
1.3(ii) Make stakeholders aware of their interactions with AI systems.Every executed response carries an AI disclosure block and a response header.Showing it to the people concerned, and the rest of the transparency duties in 1.3.
1.3(iv) Provide information that enables those adversely affected to challenge an output.A signed receipt and audit entry bind a decision to a request id, which gives a record to point to when an output is challenged.Explanations of why the model said what it said, and the process for challenges.
1.4(a) Systems are robust, secure and safe throughout their lifecycle, including under misuse.Signed, single-use, request-bound artifacts, replay protection and fail-closed behaviour.Model robustness, safety evaluation and security of the rest of the system.
1.4(b) Mechanisms so that systems can be overridden, repaired or decommissioned safely.A global kill switch and revocation, with an audited operator, reason and time.Repairing or retiring the system.
1.5(a) Accountability for proper functioning and for respecting the principles.Records that a third party can verify support, but do not replace, an accountability process.Assigning accountability and answering to it.
1.5(b) Traceability, including datasets, processes and decisions made during the lifecycle.Each decision is traceable through a request hash, an audit entry and a transparency-log leaf.Datasets and training or development traceability.
1.5(c) A systematic risk management approach on an ongoing basis.Runtime output (blocked and allowed counts and reasons) that feeds a risk process.The risk management approach itself.

Cross-check: the Regulatory Alignment Matrix, section 8

How it fits

What the OECD AI Principles asks, and what qodiqa produces.

  1. Traceability (1.5(b)) is supported by Request hash and audit entry (Decision to request).
  2. Override and decommission (1.4(b)) is supported by Kill switch and revocation (Audited operator and reason).
  3. Awareness of AI (1.3(ii)) is supported by AI disclosure block (On every executed response).

Evidence

Evidence you can hand to an auditor.

Five artifacts, each checkable on its own

  • Consent artifact Signed with Ed25519, bound to one request, single use.
  • INTENT entry The audit entry written before the provider call.
  • Receipt A signed decision record that checks offline.
  • Audit export Hash-chained entries with a Merkle inclusion proof.
  • Anchor proof A signed checkpoint pinned outside the gateway.
Verify offline, from a clone of qodiqa-verify
node bin/verify.bundle.external.js \
  --keys examples/front-half/keys.json \
  --transparency-keys examples/front-half/transparency-keys.json \
  examples/front-half/bundle-with-proof.json

The example files ship with the repository and are signed with throwaway keys. The command exits 0 and prints "valid": true. For your own evidence, take keys and anchors from a source other than the gateway that wrote it.

Scope and limits

Scope and limits

Technical evidence, not a compliance outcome

qodiqa produces verifiable technical evidence that supports parts of the OECD AI Principles. It is not a compliance certification, it is not legal advice, and no regulator or standards body has endorsed it. Whether your use meets the OECD AI Principles is a decision for you and your advisers.

What qodiqa does not address:

  • Inclusive growth, human-centred values and fairness in how you design and use AI.
  • Environmental and societal impact, and international co-operation.
  • Whether recorded decisions were correct: a receipt proves they were recorded.
  • Only requests that pass through the gateway are checked, and a receipt proves a decision was recorded, not that it was correct.

Sources

Sources

Article, clause and control references were read in the official texts listed first. The remaining links are qodiqa’s own documents. Checked on 2026-11-01; rules change, so read the current text.

Next steps

See the evidence for yourself.

Read the spec sectionRun a simulationQuickstart