DPIIT’s draft AI guidelines for TM Registry address fabricated citations but lack clarity on verification duties and proportionality. Practitioners warn of due process risks in automated evidence scrutiny.
Draft AI Guidelines for TM Registry Need Stronger Safeguards
The Trade Marks Registry has released draft guidelines on the use of artificial intelligence in proceedings, aiming to curb misuse such as submission of fake judgments, AI-generated evidence, and misleading translations. While the initiative responds to growing concerns—exemplified by recent cases involving hallucinated case law—the framework raises unresolved questions about the Registry’s verification responsibilities and due process protections.
The draft requires parties to disclose if AI tools were used in preparing submissions and mandates certification of authenticity for cited authorities. However, it remains ambiguous on whether the Registry must independently verify citations or merely rely on declarations. It also does not distinguish between AI-assisted research and fully AI-generated content, potentially penalizing legitimate use. The prescribed consequences—ranging from reprimand to rejection of filings—lack gradation based on intent or harm caused.
Further, the guidelines do not address emerging threats such as prompt injection attacks, where hidden instructions in e-filings could manipulate AI tools used by Registry officials. As highlighted in recent commentary, this vulnerability could compromise the integrity of automated legal research systems.
For practitioners, the uncertainty creates compliance risks. Until final rules clarify thresholds for disclosure, verification, and proportionality, lawyers must maintain rigorous internal validation of AI-generated content. The guidelines, if implemented as drafted, may require careful scrutiny of procedural fairness, especially in opposition and rectification proceedings.
