Safe, proportionate professional use
Safety & governance
AI can be useful without being harmless or authoritative. Build confidentiality, verification and human responsibility into the task from the start.
Use the decision toolThe two rules
AI assists.
People decide.
Never assume AI is correct because it sounds confident.
Pause before entering information
- Is the platform approved?
- Is the task suitable for AI?
- Have identifying details been removed?
- Can the answer be checked?
- Who owns the final decision?
Six practical safeguards
Govern the workflow, not just the tool
- Protect information
- Never enter identifiable client information into an unapproved AI system. Data can identify someone through a combination of details, even without their name.
- Check the answer
- AI may invent facts, references and quotations. Open sources, confirm that they exist and check that they support the claim.
- Recognise bias
- Training data and prompts can reproduce stereotypes or exclude important experience. Ask whose perspective is missing and inspect the language used.
- Keep human oversight
- A named person should understand the task, review the output and remain accountable for what is used, shared or decided.
- Follow approved use
- An individual account and an organisation-approved enterprise system may handle data differently. Policy and information-governance advice come first.
- Be transparent
- Where AI has materially shaped published or professional work, use proportionate transparency and retain a useful record of the human review.
01
02
03
04
05
06
Interactive decision aid
Should I use AI for this task?
Use this short educational check before you begin. It cannot replace your organisation’s policy or professional advice.
Educational result
Answer each question
This tool gives educational guidance, not organisational approval.
Make safe use repeatable