Tools, terms and starting point.

The Charity Commission is clear that AI does not remove or dilute trustee duties. Trustees remain responsible for decisions, including where an AI tool has supported the work. Use this as the starting point for governance conversations, alongside the charity’s own purposes, duties and risk management.

The Commission’s AI material is a blog post rather than a dedicated statutory AI guide.

• Charities and Artificial Intelligence
• Charity Sector Risk Assessment

The Fundraising Regulator’s guidance is an important reference point for any charity using or advising on AI in fundraising. It asks organisations to remain responsible and accountable for AI outputs. It covers inaccurate or generated information, hallucinations, bias, personal and non-public data, data storage, transparency and human oversight.

• Guidance on using artificial intelligence in fundraising

The ICO is the UK data protection regulator. Its AI resources matter wherever personal data, sensitive information, profiling, automated decision-making or high-risk processing is involved.

The ICO is updating its automated decision-making and profiling guidance following the Data (Use and Access) Act 2025. The consultation closed in May 2026 and the ICO lists final guidance for winter 2026.

• Guidance on AI and data protection
• Automated decision-making and profiling
• Data Protection Impact Assessments
• ICO technology plans and updates

The UK does not have one single, overarching AI law. Existing legal duties and regulators apply according to the context, particularly where AI affects recruitment, assessment, eligibility, prioritisation, risk or access to services.

The toolkit is a useful starting point for people designing, operating, procuring or delivering AI-enabled products. It was published on 8 September 2026.

This Joint Committee on Human Rights report is useful context for consequential decisions about people. It makes recommendations to government. It is not current law or government policy.

• House of Commons Library: AI regulation in the UK
• Government AI Risk Management Toolkit
• Parliament: Human rights and the regulation of AI

What it is: A conversational AI chatbot and assistant that can answer questions, write text, and process documents.

Charity example: Drafting a first version of a routine volunteer update email or summarising a lengthy public report.

What to check first: Check answers for hallucinations, review for confident but inaccurate facts, and ensure no personal or confidential data is included.

What it is: An AI assistant that can generate text, analyse documents, and help with coding tasks.

Charity example: Reviewing draft policy documents for clarity or extracting themes from anonymised consultation responses.

What to check first: Verify the context and details, ensure the tone matches your charity values, and do not input personal data.

What it is: An AI assistant built into Microsoft systems that can search the web and interact with documents.

Charity example: Drafting a meeting agenda based on publicly available information.

What to check first: Check which specific product, account and access is approved by your organisation. Do not assume work account protections are active by default.

What it is: Google's conversational AI model capable of reasoning across text, code, and images.

Charity example: Brainstorming potential event names or structuring a draft project timeline.

What to check first: Verify factual claims against reliable sources. Ensure you are signed into an enterprise account if handling sensitive work.

What it is: An AI research assistant that grounds its answers solely in the documents you upload.

Charity example: Uploading your own published strategy documents to quickly query them or generate FAQs based strictly on the text.

What to check first: Ensure you have the right to upload the documents. Check that it hasn't missed nuances in your original text.

What it is: An AI-powered notepad for meetings that drafts notes from your conversation. Granola is UK GDPR compliant.

Charity example: Capturing action points and decisions during an internal team meeting without distracting the host.

What to check first: Check whether the tool joins as a visible participant or records audio locally. Ensure attendees are informed.

What it is: An AI tool for creating presentations, documents, and webpages rapidly.

Charity example: Producing a quick slide deck outline for a training session from a list of bullet points.

What to check first: Review the layout, images, and content for accuracy and tone. Do not include confidential data in the prompt.

What it is: A voice dictation AI that turns speech into polished writing.

Charity example: Dictating quick notes or a draft email while on the move or for accessibility purposes.

What to check first: Check the transcription for misunderstood words and ensure dictation is done in a private space if discussing sensitive topics.

Artificial Intelligence. Computer systems designed to perform tasks that typically require human intelligence.

Example: Using a tool to automatically sort incoming queries or transcribe meeting audio.

A type of AI that can create new text, images, or audio based on patterns it has learned.

Example: Asking a tool to write a draft thank-you letter to donors.

The technology behind ChatGPT, Claude, Gemini and Copilot. It predicts likely words based on training patterns, but does not understand charity values or responsibilities like a person.

Example: Generating a plain-English summary of a complex regulatory document.

The instruction or question you give to an AI tool to tell it what to do.

Example: Typing 'Summarise this 10-page strategy into 3 bullet points suitable for volunteers.'

The text, image, or answer the AI produces in response to your prompt.

Example: The draft email the tool creates for you to review.

When an AI produces a confident but inaccurate or completely invented answer. These always need checking.

Example: An AI inventing a non-existent law when asked about charity regulations.

Unfair preferences or prejudices in AI outputs, often inherited from its training data.

Example: An AI tool using outdated or stereotypical language when describing a community you support.

The crucial step of a person checking, reviewing, and taking responsibility for an AI tool's work.

Example: A staff member reading and editing an AI-drafted report before it is sent to trustees.

Restricting an AI to only find answers from specific, approved documents you provide.

Example: Using NotebookLM to answer questions based only on your charity's internal handbook.

A conversational interface where you talk back and forth with an AI.

Example: Using a conversational interface to explore ideas for a fundraising campaign.

An AI tool integrated into software to help with everyday tasks like scheduling or writing.

Example: An AI feature in your email program that suggests a polite reply.

An AI that can work through a series of steps towards a goal, including using connected tools or taking actions. Permissions and human approval are critical.

Example: An AI configured to read incoming emails, categorise them, and draft a suggested response for staff approval.

The difference between using your own private login versus a charity-managed account. Organisational accounts often have better data protection agreements.

Example: Using an enterprise Microsoft Copilot account where data is not used to train public models.

A feature where the AI remembers details from past conversations to inform future answers.

Example: The AI remembering your charity's tone of voice guidelines from a previous chat.

A mode where the AI does not save the conversation history or use it for future training.

Example: Using temporary mode when discussing a sensitive, anonymised case study.

Ways to link an AI tool directly to your charity's other software, like databases or file storage.

Example: Linking an AI app to your Google Drive so it can search across all your documents.