Use Generative AI as a Drafting Partner, Not a Final Authority

Understand what a response is

Generative AI systems produce plausible text from patterns in training and prompt data; a confident tone does not demonstrate that a statement is true. Outputs may omit important context, misstate a source, or invent details. Treat the tool as a useful assistant for outlining, summarizing material you provide, or generating questions—not as a substitute for checking facts, professional judgment, or the original document.

Verify what matters

For medical, legal, financial, safety, or current-event claims, confirm the answer with a qualified person or primary source. Open cited links yourself and check that the quoted passage actually supports the claim. Ask the model to identify uncertainty or provide counterarguments, but remember that a second response from the same model is not independent confirmation. Keep a record of which portions were machine-generated when your workplace or school requires disclosure.

Use it responsibly

Do not paste private client records, passwords, confidential work, or another person's sensitive information into a service unless your organization has approved that use. Review an output for bias, missing voices, and errors before sharing it. If a task affects someone's rights or access to services, a human should review the decision and its evidence. NIST's risk-management work offers a broader framework for thinking about trustworthy AI systems.

Sources and further reading

• NIST: AI Risk Management Framework
• NIST: Generative AI Profile (PDF)