Appendix E — Responsible AI Use Policy

Status: Authoritative · Owner: PI · Review: Annually (fast-moving area)

E.1 1. Principle

AI accelerates judgment; it never replaces it. Members are fully responsible for everything they submit, regardless of which tools assisted.

E.2 2. Permitted uses (with care)

Coding assistance; writing structure/clarity/bilingual phrasing; literature orientation and summarization of papers you then read; workflow automation and documentation.

E.3 3. Prohibited uses

• Entering restricted, identifiable, IRB-governed, claims/pharmacy/clinical, or DUA-controlled data into any external AI tool. (Treat prompt boxes as public.)
• Presenting unverified AI output as fact.
• Letting AI fabricate, alter, or “clean” data or results (research misconduct).
• Submitting AI-assisted work you cannot understand or defend.
• Listing AI tools as authors.

E.4 4. Verification (mandatory)

Independently verify all AI-touched citations (exist + support the claim), numbers/stats, code (read + test), and facts/methods (primary sources).

E.5 5. Disclosure

• Internally: note material AI assistance in commits/notebooks/docs.
• For products: complete the AI-Use Disclosure form and apply the target venue’s required AI-disclosure language and any prohibitions.

E.6 6. AI and trainees

AI is a tutor/assistant, not a substitute for skill-building. Mentors and mentees discuss AI use openly in IDPs/one-on-ones to ensure it accelerates — rather than replaces — learning.

E.7 7. The defensibility test

Before submitting anything AI helped with: Could I explain and defend every claim, line of code, and citation as if no AI were involved? If not, the work is not finished.