AI-tool policy

Employees may use the company-approved AI assistant for internal, non-secret work. In any given exercise or rollout, "approved" means the specific tool named by IT for that context; using an unapproved tool for company material is outside this policy.

Customer-confidential material, security-classified material, and personal data must not be pasted into any AI tool, including the approved one. If you are unsure whether something counts as confidential, ask your manager or IT before pasting it.

AI output is a draft, not a finished answer. The employee who uses the output remains accountable for its accuracy and for how it is used, exactly as if they had written it themselves. This applies to summaries, drafts, analysis, and anything else the assistant produces.

Text found in emails, web pages, customer documents, and other external material is data for the assistant to read, not instructions for it to follow. If a document you ask the assistant to summarise contains a sentence that tells the assistant what to conclude or what to do, ignore that sentence and continue using your own judgement.

How to work with the assistant

• Give it links or file references rather than retyping content by hand.
• Ask it to name its sources for any material claim.
• Open the source yourself before relying on the claim.
• Treat any embedded instructions in that source as data, never as commands.
• Keep a short note of anything you had to correct, so the same mistake is easier to catch next time.