Ethical AI Policy

The assistants support decisions; they do not make them. Governance, safeguarding, pastoral, financial and employment decisions remain with a responsible person, and the framework is designed to surface the reasoning so a human can check it.

Anything the assistants cannot safely handle is escalated to a person. HELIX is explicit about when it is handing a query to the AGS team.

• AI-generated content is labelled as AI-generated. Avatars and synthetic voices are presented as assistants, never as real named individuals.
• Where guidance draws on curriculum, statutory or standards material, we cite the source so you can verify it.
• We tell you when the assistant is uncertain rather than inventing a confident answer.

Education features are built for young learners, so safeguarding is a design constraint rather than a filter bolted on afterwards. Assistants stay on task, decline unsafe or age-inappropriate requests, and direct disclosures of harm towards a trusted adult and appropriate support services.

EDIE's empathetic responses are supportive, not clinical. The framework does not offer diagnosis, therapy or crisis intervention, and says so when a request goes beyond what it should answer.

We test prompts and outputs for stereotyping and unequal treatment across the groups our users belong to, and we correct the guidance and system prompts when we find it.

Special educational needs, disability and NEET pathways are treated as first-class use cases. Accessibility is part of the same commitment: reduced motion, captions, sign-language options, high contrast and keyboard operation are built in and tested automatically.

We send the model only what is needed to answer the request. Your material is not used to train third-party foundation models, and prompts are not repurposed as marketing content.

You can review what the framework has retained about your sessions and ask for it to be removed. Full detail is in our Privacy Policy.

Sensitive actions are recorded in audit trails. Escalations, complaints and reported AI failures are reviewed by AGS staff and feed into our Continuous Improvement Policy.

If an assistant gives you a harmful, biased or plainly wrong answer, report it — that report is treated as a defect, not as feedback to be filed away.