When AI is part of our work, a person remains accountable for it.
We use AI to help people learn, create, and make sense of complexity. We do not use it to replace human judgment, hide responsibility, or make high-impact decisions about people.
Our approach is informed by leading international AI and data-governance frameworks. Where a law applies, we follow it. Where it does not, we still hold ourselves to the operating practices below.
A person makes the final call
AI can draft, analyse, suggest, and support. It does not make the final decision.
Every material AI use has a named person responsible for the purpose, inputs, review, release, and response if something goes wrong. We maintain human ownership over consequential decisions, including learning, client work, and public-facing outputs.
Informed by: NIST AI Risk Management Framework and OECD AI Principles on accountability.
Read the NIST AI Risk Management Framework →Read the OECD AI Principles →We disclose material AI involvement
If AI materially shapes an interaction, deliverable, or synthetic media, we say so in plain language.
That means people should not be left guessing whether they are interacting with an AI-assisted tool, whether content was generated or materially altered with AI, or whether a person reviewed the final result.
We choose this transparency standard because trust depends on context, not fine print.
Informed by: OECD transparency principles and, where applicable, Article 50 of the EU AI Act.
Read the OECD transparency principle →Read Article 50 of the EU AI Act →We verify before we rely
AI can be useful and still be wrong.
We check material facts, figures, citations, calculations, and recommendations before we publish or rely on them. We do not present AI-generated information as verified merely because it sounds confident.
For legal, medical, financial, or other professional questions, AI is not a substitute for a qualified professional.
We test for bias and exclusion
We look for foreseeable bias, stereotypes, uneven outcomes, inaccessible design, and harmful language, especially when AI affects learning, recommendations, or access to opportunity.
If a material concern cannot be resolved, we narrow the use, add human review, or do not proceed.
Informed by: UNESCO’s Recommendation on the Ethics of AI and OECD principles on fairness and human-centred values.
Read UNESCO’s AI ethics recommendation →Read the OECD human-centred values principle →Your data is not fuel
We use the minimum data needed for the purpose you agreed to. We do not put confidential, sensitive, student, client-controlled, or personal information into tools that are not approved for that use.
We protect privacy, limit access, and retain information only as long as it is needed.
Informed by: GDPR Article 5 principles of lawful, fair, transparent, purpose-limited, and data-minimised processing, where applicable.
Read GDPR Article 5 →We manage risk before release
Before a material AI use goes live, we identify what could fail: fabricated information, security exposure, privacy leakage, misuse, over-reliance, bias, and unintended action.
The higher the stakes, the stronger the controls. High-risk uses require documented testing, human approval, monitoring, and a clear way to pause or stop the system.
Citing these frameworks does not imply endorsement by the issuing bodies.
