Transparency and documentation

The last habit is the least technical and the most protective: being able to say, honestly and specifically, how AI was involved in your work.

When to disclose

• When someone is talking to a machine, they should be able to know. The EU AI Act contains transparency duties along these lines: AI systems interacting with people must generally make that apparent, and providers of generative systems must mark synthetic content in machine-readable form, with exceptions in both cases.
• When authorship matters: reports presented as your analysis, anything academic, journalistic, or evaluative. "Drafted with AI assistance, reviewed and verified by me" is a perfectly professional sentence.
• When asked. Your organisation, your client, or an auditor may ask whether and how AI was used. "I don't remember" is the answer to avoid.

What to keep

For consequential AI-assisted work, keep a lightweight trace: which tool, roughly what input, what you verified, what you changed. A sentence in the document history or a note in the project file is enough. This is the difference between "the AI made a mistake" and "here is my process, the error entered at this step": the first sounds like an excuse, the second is a professional incident report.

That closes the toolkit: sanctioned tools, the paste test, tiered verification, human-owned decisions, honest disclosure, and a trace. The final chapter compresses everything into a checklist and points you to the knowledge check.