When FDA inspected Applied Therapeutics in late 2024, a third-party vendor had deleted electronic audit trails for all 47 trial participants just two days after the inspection was announced. The outcome: a complete response letter and shareholder litigation. That case underscores why audit trails matter, and the stakes are rising as AI enters clinical workflows. Under 21 CFR Part 11, any system creating or modifying electronic records must maintain secure, timestamped logs of who did what and when. ICH E6(R3), finalized in January 2025, now explicitly covers automated actions, meaning if an AI tool generates protocol text or modifies a data field, that action requires the same audit trail as a human edit. FDA's January 2025 draft guidance introduced a seven-step credibility framework for AI in regulatory submissions, while the FDA and EMA joint principles from January 2026 call for traceable documentation of data provenance and processing steps. The EU is drafting new GMP annexes on AI, and the EU AI Act mandates automatic logging for high-risk systems. The message is clear: if you can't show exactly what your AI produced, on what inputs, and who reviewed it, you may not be inspection-ready.