Learning outcomes
- Explain how to treat fluent output as unverified until checked.
- Complete the guided activity safely.
- Check that you can know when not to rely on an answer.
Privacy, Safety and Responsible Use
Accuracy, Hallucinations and Verification
Learn to treat fluent output as unverified until checked and know when not to rely on an answer in this beginner-friendly ChatGPT lesson.
Before you begin
Use sample information. Follow your organisation's AI, privacy, copyright, security and records-management policies.
Step-by-step instructions
- Open ChatGPT Settings and Data Controls and confirm you are using the intended account or workspace.
- Ask for assumptions and verify three factual claims.
- Pause and check privacy, permissions, source material, feature availability and any instruction that affects the task.
- Review the result carefully and confirm that you can know when not to rely on an answer.
- Save only the useful, verified output and record one improvement for your next prompt.
Ian's practical tip: Start with fictional or non-sensitive information. Treat the first response as a draft, then verify facts, sources, calculations and permissions before using it.
Practical activity
Ask for assumptions and verify three factual claims. Your expected result is to know when not to rely on an answer.