HC Went to the Library to Verify: GST Show Cause Notice Quashed After AI Hallucinations Found

Jeetmal Choraria v. Union of India — Delhi High Court, November 2025

Hallucinations & Sanctions | India — Delhi HC | GST

① CASE AT A GLANCE

② BACKGROUND

Jeetmal Choraria, a taxpayer registered under the Goods and Services Tax (GST) regime, received a Show Cause Notice (SCN) from the GST enforcement authority alleging tax evasion and demanding tax, interest, and penalty. The SCN — a formal document initiating adjudication proceedings that can result in significant tax liabilities — contained multiple judicial citations to support the legal positions asserted by the authority.

When Choraria’s counsel challenged the SCN before the Delhi High Court, a detailed examination of the cited judgments was undertaken. What emerged was extraordinary: one of the cited judgments did not exist anywhere in the Indian legal system — not in the Supreme Court, not in any High Court, not in any tribunal. Another citation referenced a real case but with materially wrong details — incorrect court attribution, wrong year, or wrong party name. In short, the authority had cited an imaginary case and a distorted one.

Justice Prathiba M. Singh, one of India’s most technologically alert judges and author of several landmark technology law judgments, recognised immediately the significance of what was before the court. Before proceeding, the bench made an unusual but telling step: it directed that physical law books be retrieved from the Delhi High Court library to independently verify the citations. The books confirmed the petitioner’s case — the citations were either fabricated or materially wrong.

③ THE AI ISSUE

The core legal question: Can a GST Show Cause Notice — a foundational document in the adjudication chain that determines a taxpayer’s liability — be sustained when the legal authority cited in the notice is fictitious or materially misstated?

The secondary question, which the Delhi HC addressed head-on: When the court has reason to suspect AI-generated hallucination in an official document, what verification standard applies? The HC’s answer — physically checking the books — was a deliberate signal: the court will not take citations at face value simply because they appear in an official government document.

④ WHAT THE COURT DECIDED

• The Delhi HC set aside the GST Show Cause Notice — finding that a proceeding initiated on the basis of non-existent and misstated legal authority is a nullity [goes to the root of the adjudicatory jurisdiction].
• The HC held that a Show Cause Notice must cite accurate legal authority; a notice that cites a judgment that does not exist deprives the noticee of the ability to respond to the legal basis of the proceeding — a violation of the right to a fair hearing [audi alteram partem].
• The court’s use of physical library verification was an explicit judicial statement: AI-generated citations will be verified against primary sources, and courts will do so themselves if necessary. The era of citing unverified legal authority — whether generated by AI or otherwise — is over.
• The bench specifically noted that the GST authority’s reliance on unverified AI-generated citations was not merely a procedural lapse — it demonstrated that the entire reasoning process underlying the SCN was defective.
• The matter was remitted for fresh consideration with a direction that any fresh SCN must cite only verified, genuine legal authority and must follow the principles of natural justice scrupulously.

⑤ KEY QUOTE FROM THE JUDGMENT

“We physically retrieved books from our library to verify the citations. One judgment does not exist. Another citation is wrong. A Show Cause Notice built on fictitious legal authority is no notice in the eyes of law.”

— Justice Prathiba M. Singh, Delhi High Court, WP(C) 16754/2025, 18 November 2025

⑥ THE INDIA ANGLE

Indian Law Equivalent

The Jeetmal Choraria judgment is significant across multiple Indian legal frameworks:

• Section 73/74 of the Central Goods and Services Tax Act, 2017 (CGST Act) — The Show Cause Notice under Sections 73 and 74 is the gateway to GST adjudication. A legally invalid SCN terminates the entire proceeding. The HC’s ruling follows this principle.
• Article 14 and 21 of the Constitution — The right to a fair hearing before a quasi-judicial authority is a fundamental right. An SCN that cites non-existent law deprives the noticee of the ability to understand and respond to the charge — a constitutional violation.
• Principles of Natural Justice (Audi Alteram Partem) — The noticee must know the real legal basis of the proceeding against them. Fabricated citations defeat this fundamental requirement.
• The Customs Excise and Service Tax Appellate Tribunal (CESTAT) and GST Appellate Tribunal (GSTAT) — As the GST appellate infrastructure develops, the Choraria principle is likely to be applied in tribunal proceedings challenging SCNs.

Bar Council of India Rules

While GST officers are not advocates, the Choraria case has direct implications for tax advocates appearing before GST authorities. An advocate who assists in drafting a GST SCN, representation, or reply — and who uses AI to generate citations without verification — faces the same professional risk as an advocate filing fake citations before a court.

Tax practitioners should note that Justice Prathiba M. Singh is among the most prominent voices in Indian courts on AI and law. Her bench’s decision to physically verify citations was a deliberate judicial signal, and further AI-related orders from her court are likely to be equally exacting.

Practical Advice for Indian Advocates

• In any GST or indirect tax matter — at the SCN stage, adjudication stage, or appellate stage — request from the authority a copy of each cited judgment in full. If a judgment cannot be produced, challenge its authenticity immediately and cite Jeetmal Choraria as authority for the proposition that a notice built on fictitious citations is void.
• When receiving an SCN or any quasi-judicial notice that cites case law, verify every citation before responding. If any citation is false, raise it as a preliminary objection in your reply — and preserve that objection for appeal.
• For advocates drafting SCNs, representations, or orders: verify each citation from Manupatra, SCC Online, or IndianKanoon before use. Screenshot or save your verification — it creates an audit trail if the citation is later challenged.

— — —

DEEP DIVE: The Judge Who Went to the Library — and What She Found

A GST Notice That Could Not Withstand Basic Scrutiny

The GST regime, introduced in India in 2017, has created an enormous volume of enforcement proceedings: Show Cause Notices, demand orders, penalty orders, and adjudication proceedings at various levels. Enforcement officers across India routinely draft complex SCNs that must set out the factual basis for the demand and the legal authority for the proposed action.

As AI tools have become more accessible, it is reasonable to infer that enforcement officers — like lawyers, judges, and tax authorities across India — have begun using them to assist with legal research and drafting. The Jeetmal Choraria case revealed the consequences when that use is unverified: an SCN that cited a judgment that never existed and another that was materially misstated.

When the matter came before the Delhi HC’s Division Bench, Justice Prathiba M. Singh did not merely accept the petitioner’s submission that the citations were wrong. She directed physical verification. This was a judicial act of extraordinary significance: in the digital age, a High Court judge ordered physical law books to be produced from the court library to check citations in an official government document. The books confirmed what counsel had submitted. One citation was entirely fictitious. Another was wrong in material particulars.

Justice Prathiba M. Singh and the Emerging Indian AI Law Jurisprudence

Justice Prathiba M. Singh is not an unfamiliar name in Indian technology law jurisprudence. As a judge of the Delhi High Court, she has authored landmark decisions on intermediary liability, copyright in the digital age, and data protection. Her approach in Jeetmal Choraria reflects a sophisticated understanding of how AI hallucinations work and why they are particularly dangerous in legal proceedings.

An AI hallucination is not merely an error — it is a confident, plausible-sounding falsehood that appears as authoritative as genuine content. In a legal document, this is especially dangerous because the reader has no way to distinguish a real citation from a fabricated one without verification. The HC’s response — verify against physical books — was a message not just to the GST authority in this case but to every enforcement officer, advocate, and court officer in India: AI output is not evidence; verification is mandatory.

The Choraria judgment follows a clear pattern in India’s emerging AI law jurisprudence. It joins the KMG Wires (Bombay HC, 2025) case — where a Rs. 27.91 Crore income tax assessment was quashed for the same reason — and the Gummadi Usha Rani (AP HC + SC, 2026) case — where a judge’s own use of AI hallucinations was treated as judicial misconduct. Together, these cases establish a clear principle: in India’s courts and quasi-judicial proceedings, there is no room for unverified AI-generated legal authority.

The GST Enforcement Risk: Systemic Implications

India has thousands of GST enforcement officers, many of whom handle large caseloads. The pressure to draft SCNs quickly, across complex factual and legal matrices, creates an obvious temptation to use AI tools. If even a fraction of those SCNs contain AI-hallucinated citations — citations that are never challenged because taxpayers settle or do not engage proper legal counsel — the scale of the problem is enormous.

The Choraria case now gives every GST noticee a procedural tool: request verification of every citation in the SCN. If even one citation is false, the entire notice may be void. This could have significant implications for GST enforcement revenue, and the GST Council and CBIC would be wise to respond with immediate guidance to officers on AI use.

What to Watch For

Justice Prathiba M. Singh’s bench is likely to continue developing this area of law. Her court has heard multiple AI-related matters and has demonstrated a willingness to take technologically sophisticated positions. Any further SCN challenges based on AI hallucinations in her court will likely receive equally careful scrutiny.

The Supreme Court’s proceedings in Gummadi (SLP 7575/2026), which involve notices to the AG, SG, and Bar Council, may also address the quasi-judicial context — potentially issuing directions that apply to GST officers and other administrative authorities as well as to advocates and judges.

For taxpayers and their advisors, the immediate message is clear: a GST SCN that cites fake judgments is a nullity. The Delhi HC has established this; the tool to challenge it is already in advocates’ hands.