AI Ethics & Bias Mitigation

Generative tools reproduce the assumptions in their training data. Left unchecked, that shows up in examples, images and scenarios. This policy is how we catch it before a room full of people does.

What it covers

• AI-assisted examples, case studies, scenarios, images, exercises, summaries and learner-support interactions.
• Recurring AI-enabled platform workflows, and any material change of model, provider, prompt architecture or use case.
• Tool and vendor approval, where published fairness, safety or responsible-AI information is relevant to the intended use.

What we commit to

• AI-supported output is reviewed for stereotypes, exclusion, inappropriate assumptions, unbalanced representation, language neutrality, accessibility and culturally inappropriate framing.
• Review considers sex and gender, nationality, race and ethnicity, religion, disability, age and socioeconomic assumptions, alongside the cultural context of delivery in the region.
• A formal bias and fairness audit is completed at least quarterly for recurring AI-enabled workflows, and after any material model or system change.
• Learner feedback is read for signs of bias, exclusion or cultural inappropriateness that design review did not catch.
• Tool approval takes account of published responsible-AI, safety and fairness information. We do not claim knowledge of a vendor's training data when the vendor does not disclose it.
• Identified bias is corrected, logged and fed back into content, prompts, examples, tool restrictions or trainer guidance. A clean review is recorded too, as evidence the check happened.
• AI is never the sole decision-maker on assessment, grading, completion or access.

What this means for you

The examples in the room should recognize the people in it. If they do not, tell us; that feedback goes into the audit rather than into a drawer.