On 2 September 2026 the Supreme Court set aside the ₹425.28 crore customs penalty imposed on Vijay Ghanshyam Gadiya. An administrative order cannot rest on AI‑produced case‑law references that are unverified. The penalty against the diamond trader is annulled and the matter will be reheard by a different customs officer. Reliance on hallucinated AI output defeats the validity of statutory adjudication.

Full News Breakdown

The dispute originated from a penalty order that alleged the appellant misdeclared natural diamonds as lab‑grown, with the adjudicating officer supporting the finding by citing several judgments and articles generated by artificial intelligence. The appellant challenged the authenticity of those citations, leading the apex court to examine the veracity of the AI‑derived material. Verification revealed many references to be non‑existent or mis‑applied, resulting in the penalty being set aside.

• Case Name: Vijay Ghanshyam Gadiya v. Union of India and Another

• Court: Supreme Court of India

• Bench: Justices Dipankar Datta and Sheel Nagu

• Date: 2 September 2026

• Statutes Cited: Customs Act, 1962

• Key Provisions: Section 114

• Primary Legal Issue: Whether an administrative penalty can be sustained on the basis of AI‑generated, non‑existent case‑law citations

• Petitioner Arguments: The cited judgments and articles were produced by AI and were either non‑existent or mis‑interpreted, rendering the penalty unsustainable

• Respondent Arguments: The customs officer relied on those citations to justify the penalty, asserting they were valid precedents

• Court's Reasoning: Verification revealed many citations were fictitious or incorrectly applied; reliance on such material defeats the legitimacy of the order

• Ratio Decidendi: An administrative authority may not base a penalty order on AI‑generated citations that are unverified or fictitious; such reliance is fatal to the order’s validity

• Operative Order: The penalty order and the confirming Gujarat High Court judgment are set aside; the matter is remitted for fresh adjudication by an officer of the same rank other than the original adjudicating officer

• Practical Outcome: The appellant’s penalty is annulled and the case will be reheard; the appointing authority may consider disciplinary action against the officer who authored the flawed order

How Does This Affect You?

Before the judgment, practitioners faced uncertainty about the evidentiary weight of AI‑produced case‑law references in customs penalty proceedings. The Court resolved that reliance on unverified AI output renders an administrative order voidable. Parties now have a clear basis to challenge penalties that rest on such material, and authorities are obliged to verify the authenticity of every cited authority before it can support a sanction. The shift creates greater certainty for appellants while imposing heightened verification duties on officials drafting penalty orders.

For Lawyers & Advocates

• Pending customs penalty appeals provide a basis for applications for review on the ground that the order relies on unauthenticated AI‑derived references, invoking Section 114.

• Pleading templates that incorporate a verification clause, requiring a certified copy of every cited judgment, reduce the risk of reliance on hallucinated outputs.

• Preserving electronic search logs, AI‑tool audit trails, and screen‑captures of research queries creates a record that courts can scrutinise for provenance of cited material.

• Orders authored with unverified AI citations expose senior customs officials to disciplinary action, prompting internal compliance audits of research practices.

• Treating AI‑generated summaries as secondary material and citing only binding Supreme Court or High Court judgments with proper citations satisfies the natural‑justice requirement.

For Law Students

This case teaches that courts will not tolerate reliance on artificial‑intelligence outputs that cannot be authenticated when exercising regulatory power. The core doctrine is the principle of natural justice demanding decisions be based on genuine, verifiable legal authorities.

Comparable cases include Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd, 2025 SC, which warned against unverified AI precedents, and the NCLT judgment set aside in 2026 for reliance on hallucinated AI citations, illustrating the judiciary’s evolving stance on AI‑driven research.

For Businesses

• Manual verification of all legal authorities cited in customs clearance filings reduces the risk of penalty reversal and subsequent re‑assessment for diamond importers.

• Embedding a verification step in standard operating procedures for drafting penalty orders mitigates the risk of disciplinary measures for officers who incorporate fictitious AI references.

• Disclosing the provenance and verification status of case‑law outputs limits liability for AI‑legal‑tech providers who facilitate hallucinated citations that influence regulatory decisions.

• Board‑level risk committees that assess the firm’s reliance on AI tools for legal research and implement policies requiring human validation of any citation used in regulatory submissions strengthen governance.

Key Takeaways

• Administrative orders that rely on unverified AI‑generated citations lack authenticity and fail to meet the statutory requirement.

• Embedding a verification step for every cited authority in customs penalty matters and preserving audit trails of AI research aligns with the Court’s requirement.

• Regulators and adjudicating officers can no longer justify sanctions with hallucinated AI output, narrowing the evidentiary scope of such material.

• The Ministry of Finance is expected to issue guidelines on AI‑assisted legal research that will formalise verification protocols by early 2027.

• In‑house counsel reviewing AI‑driven legal research workflows before the next fiscal year will ensure compliance with the new judicial standard.

Source: Supreme Court scraps ₹425 crore penalty over fake citations, AI hallucinations in customs order