AI Generated Documents Provenance And Detection

A I - GENERATEDDOCUMENTS · AGUIDEFORUKLAWYERS AI-Generated Documents: Provenance and Detection Fabricated Evidence, Invented Citations and the Return to Provenance COMPUTERFORENSICSLAB § ABOUTTHEAUTHOR PRE PA REDBYCOMPUTERFORENSICSLABE - DISCOVERYTEAM ESTABLISHED 2007 · LONDON ISO 17025-ALIGNED PROCEDURES DOCUMENT PROVENANCE ANALYSIS § CONTENTS In this guide 01 Executive summary 02 The problem in plain English: plausible text is cheap now 03 How AI-generated and AI-assisted documents arise 04 Why text-based detection is weak 05 Provenance: the forensic answer 06 Metadata, corroboration and honest limits 07 Deployment: fabrication, hallucinated citations and duties 08 Source architecture: where else the evidence lives 09 Worked examples 10 Common mistakes and technical limitations 11 Questions to ask · Suggested wording 12 Checklist and red flags · When to involve a digital forensic expert 13 Frequently asked questions 14 Glossary · References · Disclaimer · How a specialist laboratory can assist § 0 1 · ORIENTATION Executive summary The headline point: AI can now fabricate a plausible document in seconds, but detecting AI from the text alone is unreliable, so the forensic answer is not a detector but provenance: where the document actually came support its claimed origin. § 0 2 · FIRSTPRINCIPLES The problem in plain English: plausible text is cheap now § 0 3 · HOWTHEYARISE How AI-generated and AI-assisted documents arise § 0 4 · WHYTEXTDETECTIONISWEAK Why text-based detection is weak § 0 5 · PROVENANCE Provenance: the forensic answer § 0 6 · METADATAANDHONESTLIMITS Metadata, corroboration and honest limits § 0 7 · DEPLOYMENT Deployment: fabrication, hallucinated citations and duties § 0 8 · THEWIDERMAP Source architecture: where else the evidence lives CORROBORATING RECIPIENT / SERVER / BACKUPS / APPLICATION / SOURCE COUNTERPART CLOUD VERSIONS DATABASE THE SENDER / DOCUMENT AUTHOR ITSELF SYSTEM § 0 9 · INTHEWILD Worked examples EXAMPLE1 · THEEMAILTHATEXISTEDONLYASAFILE EXAMPLE2 · THESUBMISSIONWITHAUTHORITIESTHATDIDNOTEXIST EXAMPLE3 · THEGENUINEDOCUMENTDEFENDEDBYITSPROVENANCE § 1 0 · WHEREITGOESWRONG Common mistakes and technical limitations Common mistakes Technical limitations § 1 1 · INTERROGATORIES & DRAFTINGAIDS Questions to ask · Suggested wording Ask your client Ask your opponent Ask your eDiscovery / forensic provider SUGGESTEDWORDING · INSTRUCTIONFORADOCUMENT- P R OV ENANCEEXAMIN AT ION § 1 2 · QUICKCONTROL Checklist and red flags · When to involve a digital forensic expert The provenance checklist Red flags When to involve a digital forensic expert § 1 3 · COMMONQUESTIONS Frequently asked questions Can you tell if a document was written by AI? How do you prove a document is fabricated? Why do we need the native file and not a printout? Our opponent's submissions cite cases we cannot find. What does that mean? Can metadata itself be faked? Is using AI to draft documents a problem in itself? § 1 4 · REFERENCE Glossary Sources and authoritative references 19 reporting): cflab.uk/digital-forensics-services · guides library: cflab.uk/guides DISCLAIMER § HOWASPECIALISTLABORATORYCANASSIST Working with Computer Forensics Lab Speak to a forensic examiner, not a salesperson. INSTRUCTTHELAB NEWENQUIRIESEMAILE - DISCOVERY