Seven principles every member stands behind.

Seven principles, one paragraph each. No legalese, no footnotes, no hidden carve-outs. Read them and you know exactly what an IoAI member has agreed to.

1. Principle I

   Integrity

   Members are honest about what AI can and cannot do, what they know and do not know, and where their work begins and ends. We do not oversell, ghost-write our credentials, or hide our limits.

   • Truthful claims with sourced numbers, including the model and method behind any benchmark result
   • Disclosure when generative AI shaped a piece of work that carries our name
   • Mistakes corrected in writing, with the same reach as the original claim

   “I will not say more than I know.”

2. Principle II

   Responsibility

   Members own the AI systems they design, deploy, or recommend — through to the consequences for the people on the other side of the screen. Outsourcing the model does not outsource the duty.

   • A named human accountable for every production decision a member ships
   • A written record of who signed off on which release
   • Rollback plans designed before launch, not after the first incident

   “I will own the outcomes of what I build.”

3. Principle III

   Fairness

   Members work to ensure AI systems treat people equitably across protected characteristics, postcodes, and life circumstances. We do not let convenient datasets quietly reproduce historical harm.

   • Bias tested, documented, and mitigated before release — and re-tested as the world changes
   • Disaggregated performance metrics in every model card
   • No system ships whose error rate falls disproportionately on a protected group

   “I will measure who pays the cost of my errors.”

4. Principle IV

   Privacy

   Members protect the personal data passed through the systems they touch. We collect what is necessary, hold it as briefly as the work allows, and never repurpose it without consent.

   • Data minimisation as the default, not an audit-time scramble
   • Synthetic or anonymised data wherever the use case permits
   • A documented retention schedule for every dataset under our care

   “I will treat the data I touch as if it belonged to my family.”

5. Principle V

   Safety

   Members do not ship AI systems whose failure modes have not been thought through. We red-team, we monitor, and we hold the launch when the evidence is not yet there.

   • A pre-launch adversarial review for any user-facing model
   • Live monitoring of safety-relevant signals, with thresholds agreed in advance
   • A standing right to delay release without commercial penalty

   “I will hold the launch if the evidence is not there.”

6. Principle VI

   Continuous learning

   AI changes quickly, and so must the people working in it. Members keep their practice current and share what they learn with peers, juniors, and the public.

   • Recorded continuing professional development every accreditation cycle
   • At least one public contribution a year — a talk, a paper, a clinic, a mentoring relationship
   • Honest peer review, given and received

   “I will keep learning, and I will pass it on.”

7. Principle VII

   Public interest

   When a commercial duty conflicts with the safety, dignity, or autonomy of the public, the public interest comes first. The credential is for service, not insulation.

   • Members speak up — internally first, externally if required — when a system risks meaningful harm
   • Whistleblower support through the IoAI Standards Council
   • Support for colleagues who raise concerns in good faith

   “I will put the public ahead of the brief, when the brief is wrong.”

Signed at enrolment, re-affirmed at every annual renewal.