Purpose
This policy sets out how The Amber Trust will use artificial intelligence (AI) responsibly to support our work.
Definition
When we refer to AI in this document, we mean AI tools based upon generative large language models, such as Chat-GPT, Claude, Gemini, etc.
Principles
Supportive, not substitutive: AI may be used to assist with administrative tasks, communications, or resource creation, but it will never replace the human relationships at the heart of our work.
- Accessibility first: Any AI tools we use must support accessibility and not create barriers for children, families, staff, or volunteers.
- Transparency: We will be open about when and how AI is used in our activities and materials.
- Human oversight: AI outputs will always be reviewed by a staff member or Trustee before being shared publicly or used in decision-making.
- Privacy and data protection: AI will only be used in line with UK GDPR and data protection law. We will not input sensitive personal information into AI tools unless we are sure they are secure and compliant.
- Ethical use: We will avoid using AI in ways that could cause harm, spread misinformation, or disadvantage the children and families we serve.
Risks Associated with AI
Legal and Compliance Risks
- Data protection breaches: putting personal or sensitive information e.g. children’s health or family details, into AI tools could breach UK GDPR.
- Unclear accountability: if AI generates misleading or inappropriate material, The Amber Trust is still responsible.
- Copyright issues: some AI outputs may reuse or imitate copyrighted content.
Safeguarding and Ethical Risks
- Replacing human judgement: relying on AI for decisions about children’s learning, needs, or welfare could be unsafe.
- Dehumanisation: children and families may feel less valued if AI replaces personal communication or creativity.
- Bias and discrimination: AI tools can reflect or amplify harmful stereotypes e.g. about disability, culture, or background.
Accuracy and Quality Risks
- Inaccurate or misleading information: AI can produce answers that sound plausible but are wrong.
- Inappropriate tone: AI-generated text may lack sensitivity, warmth, or inclusivity.
- Over-reliance: staff and volunteers might stop developing their own skills if they lean too heavily on AI.
Security and Trust Risks
- Data storage uncertainty: many AI tools send data outside the UK/EU, raising risks about where information is stored.
- Loss of trust: families may feel uncomfortable if they find out AI was used without transparency.
- Reputation damage: using AI poorly e.g. publishing unchecked content, could harm the charity’s reputation.
Areas of Acceptable Use
In order to mitigate the risks identified above the following areas of acceptable use have been identified.
AI may be used for:
- Drafting or checking written materials e.g. newsletters or grant applications.
- Generating ideas for activities or resources.
- Supporting accessibility e.g. transcriptions, alt-text.
- Assisting with simple administrative tasks.
AI may not be used for:
- Making decisions about individual children or families.
- Replacing human-led teaching, mentoring, or safeguarding.
- Handling or storing sensitive personal information without explicit safeguards.
AI Quick Guide for Employees
It is acceptable to use AI to:
- Draft emails, newsletters, or grant applications.
- Check spelling, grammar, or readability.
- Create ideas for activities or events.
- Support accessibility e.g. transcriptions or alt-text.
- Make writing more concise.
- Create formulas for Excel.
However:
- Always review and edit AI’s work before sharing it.
- Be clear if something has been created with the help of AI.
Employees should never:
- Put private, sensitive, personally identifiable or commercially confidential information about children, families, staff or the charity into AI tools.
- Use AI to make decisions about children or families.
- Use AI instead of human teaching, mentoring, or safeguarding.
- Share AI output without checking it first.
If an employee is unsure whether AI usage is acceptable, they should seek advice from the CEO.
Review frequency: Once per year