Skip to content

Incident Tracker / India

Vijay Ghanshyam Gadiya v. Union of India, 2026 INSC 947

·Supreme Court of India·INOther

Extending Pooja Ramesh Singh from tribunals to executive adjudicators, the Supreme Court of India set aside a customs penalty order of INR 4.25 billion, and the High Court judgment upholding it, because the order relied on AI-hallucinated precedents. Justice Datta held even an iota of fake material vitiates a decision and left disciplinary action to the officer's appointing authority.

Analysis
India's Supreme Court sets aside customs penalty over AI-hallucinated citations
Court
Supreme Court of India
Date
2026-09-02
Who used the AI
Other
AI tool
Not identified in the record
Conduct
An Order-in-Original of the Additional Commissioner of Customs, Surat, dated 8 October 2025 and imposing a penalty of INR 425.28 crore, relied on judgments and articles that were nonexistent, carried fake citations, or did not support the propositions cited. The Gujarat High Court had dismissed the challenge.
Outcome
Other
Penalty
No monetary penalty recorded
Regulatory outcome
No separate regulatory disposition recorded
Status
Verified against listed sources
Last checked
2026-09-04

What was fabricated or misused

  • several judgments and articles cited in the Order-in-Original dated 8th October, 2025 that were generated using Artificial Intelligence — the Supreme Court did not list them
  • case laws relied upon by the customs adjudicator that were non-existent or had fake citations — the Supreme Court did not list them
  • some real case laws relied upon by the customs adjudicator that did not lay down the ratio deduced therefrom — the Supreme Court did not list them

What the court said

“Upon verifying such judgments and articles, it does appear that the second respondent has relied upon case laws that are either non-existent or have fake citations.”
Supreme Court of India, [4]
“A further investigation has also revealed that some of the case laws relied upon by the second respondent, which do exist, do not lay down the ratio deduced therefrom and it appears to be a hallucination of AI.”
Supreme Court of India, [4]
“assistance can never be substituted for adjudication. AI may well serve as training wheels but entrusting it with the pilot’s seat would be both imprudent and dangerous.”
Supreme Court of India, [6]
“the impugned order of the High Court confirming the penalty is consequently set aside together with the Order-in-Original”
Supreme Court of India, [8]

Timeline

  1. Additional Commissioner of Customs imposed a Rs. 425,27,99,100 penalty.
  2. Gujarat High Court dismissed Gadiya's challenge to the penalty.
  3. Supreme Court set aside the High Court order and Order-in-Original.

Why this case matters

Gadiya extends Pooja Ramesh Singh from tribunal judgments to an executive customs adjudication. The Supreme Court did not resolve the customs merits; it set aside a ₹425.28 crore penalty because fake and misused authorities entered the reasoning. That makes it part of India's unusual cluster in which the faulty AI material appears in the adjudicator's own order, alongside Buckeye Trust, KMG Wires, Sammaan Capital and Pooja Ramesh Singh. The remand to a different officer and the note to the appointing authority make the remedy institutional as well as case-specific.

Practice note

An adjudicator using AI as research help should keep the decision-making role with the officer. Before issuing an order, every judgment, article and quoted ratio should be checked in an official database and in the source text. If a cited case exists, verify that it supports the stated ratio; if it does not, remove or rewrite the reasoning.

Primary sources for this incident

Related regulation records for this incident

No related regulation record has been linked to this incident yet.

Read our full analysis of this case →

Similar incidents on the record

Cite this record

SafeLegalAI Global Legal AI Incident Tracker, “Vijay Ghanshyam Gadiya v. Union of India, 2026 INSC 947”, safelegalai.com/tracker/gadiya-v-union-of-india (accessed 2026-10-07). Data: CC BY 4.0.

Report an issue with this record →