Legal expert Rohit Pradhan warns courts about 'hidden AI prompts' in e-filed petitions that could manipulate AI-assisted research tools used by judges, calling for structural safeguards.
Hidden AI Prompts in Pleadings Pose Judicial Risk, Says IP Litigator
As Indian courts increasingly adopt artificial intelligence for case management and legal research, a new threat has emerged: 'prompt injection' in e-filed documents. Rohit Pradhan, an IP litigator, has highlighted the risk that embedded, invisible code in digital pleadings could manipulate AI tools used by benches to retrieve precedents or draft orders.
Unlike AI hallucinations—fabricated but unintentional outputs—this threat is deliberate and stealthy. For instance, a petition might contain metadata instructing a court’s AI system to prioritize certain judgments or suppress others. Pradhan urges that e-filing portals be audited for such vulnerabilities and that a judicial protocol be established for verifying AI-assisted outputs.
For practitioners, the warning underscores the need for transparency in digital submissions. Courts may soon require disclosure of AI use in filings, akin to conflict-of-interest statements. The issue may prompt amendments to the proposed Digital Evidence Rules under the BSA, particularly around authentication of machine-generated content.
