The Gujarat High Court has quashed a series of GST proceedings against Faiz Enterprise after finding that the tax authorities had relied on judicial precedents that were non-existent, inaccurately cited or irrelevant to the dispute. The case has emerged as a significant warning about the risks of using artificial intelligence for legal research without independently verifying the material generated by AI systems.
The dispute began with the cancellation of Faiz Enterprise’s GST registration. During proceedings, the taxpayer challenged several judicial authorities cited by the State Tax Officer, including cases that allegedly did not exist or whose citations and legal reasoning did not support the conclusions for which they were cited. The High Court subsequently found that the impugned order appeared to have relied exclusively on AI-generated case law and directed the department to reconsider the matter.
Gujarat HC Rejects GST Order Based on AI-Generated Case Law
The case involved proceedings initiated against Faiz Enterprise after the State Tax Officer issued a show-cause notice on October 30, 2025, followed by an order cancelling the company’s GST registration on December 15, 2025.
Faiz Enterprise subsequently sought revocation of the cancellation. That application was rejected in March 2026, and the taxpayer then pursued the statutory appellate remedy. The appellate order issued on May 25, 2026, also went against the taxpayer.
The dispute took a different turn when the taxpayer examined the judicial authorities relied upon by the tax officer and found serious problems with the cited cases.
Three Judicial Authorities Came Under Scrutiny
The taxpayer pointed out that several authorities cited in the GST order either could not be located or did not support the legal propositions attributed to them.
Among the authorities questioned were State of Gujarat v. Aarbee Structures Pvt. Ltd., M/s. Sri Vinayaga Agencies v. Assistant Commissioner, and Jyoti Tar Products Pvt. Ltd. v. Deputy Commissioner.
The High Court subsequently took note of the apparent problems with the citations and questioned whether the order had been generated using AI without proper verification.
What Happened in the GST Case
| Stage | Date | Development |
|---|---|---|
| GST show-cause notice | October 30, 2025 | Proceedings initiated against Faiz Enterprise |
| GST cancellation order | December 15, 2025 | Registration cancelled |
| Revocation application rejected | March 10, 2026 | Request to restore registration rejected |
| Communication of rejection | March 23, 2026 | Order communicated to taxpayer |
| Appellate order | May 25, 2026 | Appeal rejected |
| High Court intervention | August 13, 2026 | Court raised concerns over cited case law |
| Officer’s appearance | August 20, 2026 | State Tax Officer appeared and tendered apology |
| Final High Court action | August 20, 2026 | GST proceedings and related orders quashed |
The High Court ultimately accepted the department’s undertaking that the matter would be reconsidered and a fresh order would be passed in accordance with law.
Officer Apologizes After AI Use Comes to Light
The matter became particularly significant when the State Tax Officer, Devang Arvindkumar Yadav, appeared before the High Court and filed an affidavit tendering an unconditional apology.
The court had earlier indicated that the officer could be required to remain personally present if the allegations regarding reliance on non-existent or irrelevant judicial decisions were established.
The episode illustrates the difference between using AI as a research aid and treating AI-generated material as verified legal authority. A language model can produce citations that appear authentic because they follow familiar legal formatting, even when the underlying case does not exist or the cited decision does not contain the proposition attributed to it.
Why AI Hallucinations Matter in Legal Proceedings
An AI hallucination occurs when an AI system generates information that appears plausible but is factually incorrect. In legal research, fabricated case names, incorrect citations or distorted holdings can have particularly serious consequences.
AI TOOL
↓
Generates Legal Citation
↓
Citation Looks Plausible
↓
No Independent Verification
↓
Citation Included in Official Order
↓
Incorrect / Non-Existent Authority
↓
Legal Decision Becomes Vulnerable
The Gujarat case demonstrates why the final responsibility cannot be transferred to an AI system. Once an official decision is issued, the authority signing it remains responsible for the accuracy and legal reasoning contained in the document.
Gujarat HC Had Already Introduced AI Safeguards
The latest case is particularly notable because the Gujarat High Court had already issued a formal policy governing the use of artificial intelligence in judicial and court administration.
The policy recognizes that AI can improve efficiency in areas such as research and administration but warns against over-reliance, bias and erosion of human judgment. It specifically states that the fact that an AI-generated citation is correctly formatted or internally consistent does not establish that the cited authority actually exists or is accurate.
The policy also requires users to verify AI-generated material before relying on it and places responsibility on the person who authenticates AI-generated output. It states that AI use does not provide a defense against findings of error, misconduct or professional negligence.
Key Principles in Gujarat HC’s AI Policy
| AI-Related Risk | Policy Position |
|---|---|
| AI-generated case citation | Must be independently verified |
| AI-generated summary | Original document must be read before reliance |
| Machine translation | Must be verified by a competent person |
| AI-assisted transcription | Must be reviewed and certified |
| AI-generated official material | Responsibility remains with the user |
| AI-related error | Cannot be excused simply by blaming the AI system |
| Judicial reasoning | AI cannot replace human judgment |
The policy therefore establishes a clear principle: AI may assist with legal work, but accountability remains human.
Why the Case Is Important Beyond GST
The ruling has implications beyond GST administration because government departments, courts, lawyers and businesses are increasingly experimenting with generative AI.
AI tools can search large amounts of information quickly and help users summarize documents, identify potentially relevant cases and organize research. But legal systems require a higher level of reliability than ordinary information retrieval.
A fabricated citation in a casual research document may be corrected relatively easily. The same citation in a tax order, court filing or judicial decision can affect a person’s rights, financial obligations or ability to conduct business.
The Gujarat High Court case therefore provides a practical example of what can happen when AI-generated research enters a formal decision-making process without adequate human verification.
GST Registration Cancellation Was Completely Set Aside
The High Court did not simply order the tax department to correct the problematic citations. It quashed and set aside the entire sequence of proceedings challenged by the taxpayer.
This included the October 2025 show-cause notice, the December 2025 GST cancellation order, the order rejecting the revocation application and the May 2026 appellate order, along with the related GST APL-04.
The department was directed to issue a fresh notice and conduct the proceedings again. The taxpayer’s reply and defense must be considered, and the authorities must pass a reasoned order in accordance with law.
OLD PROCEEDINGS
│
├── Show-Cause Notice
├── GST Cancellation
├── Revocation Rejection
└── Appellate Order
↓
QUASHED BY HC
↓
Fresh Proceedings
↓
Fresh Reasoned Order
This means the High Court did not finally decide the underlying GST dispute in favor of the taxpayer on its substantive merits. Instead, the defective proceedings were set aside and the department was given the opportunity to reconsider the matter lawfully.
The Difference Between AI Assistance and AI Reliance
The case highlights an important distinction for government and legal professionals.
AI can be used to identify potentially relevant cases, summarize lengthy judgments or suggest legal research directions. But every citation still needs to be checked against an authoritative source before it is included in a formal document.
A safe workflow would look like this:
| Step | Recommended Practice |
|---|---|
| 1 | Use AI to identify potentially relevant authorities |
| 2 | Locate the original judgment |
| 3 | Verify the case name and citation |
| 4 | Read the relevant portion of the judgment |
| 5 | Confirm that the legal proposition actually applies |
| 6 | Independently construct the legal reasoning |
| 7 | Review the final document before authentication |
The problem in the Gujarat case was not simply that AI was used. The critical issue was the apparent absence of adequate verification and independent legal reasoning before the AI-generated authorities were relied upon in the official order.
Implications for Government Departments
The ruling could encourage government departments to introduce stronger controls around generative AI.
Tax authorities increasingly handle large volumes of legal material, making AI-assisted research attractive. However, tax orders are quasi-judicial decisions and therefore require reasoned findings based on verified law and evidence.
Departments may now face greater pressure to establish internal review procedures, particularly where officers use generative AI for legal research. Such procedures could include mandatory verification of citations, preservation of source documents and human review before AI-assisted material becomes part of an official order.
The Gujarat High Court’s existing AI policy provides a framework for this approach.
Implications for Lawyers and Businesses
The case also matters to taxpayers and legal professionals.
Businesses challenging a tax order can examine the legal authorities cited by an adjudicating officer and verify them against reliable legal databases or original judgments. If a decision relies on a non-existent or irrelevant precedent, that could become an important procedural issue in an appeal or judicial review.
For lawyers, the lesson is equally direct: AI-generated legal research should be treated as a starting point rather than a substitute for primary-source research.
The Bigger Picture
The Gujarat High Court’s decision is an early and significant example of the legal consequences of unchecked AI-generated information entering government decision-making. The court’s action shows that a citation that looks authentic is not enough; legal authorities must actually exist, say what the decision claims they say and be relevant to the dispute.
The ruling also reinforces a broader principle emerging around professional AI use: automation can assist decision-makers, but it cannot replace accountability. As government departments and legal professionals increasingly adopt generative AI, verification, source checking and human judgment will remain essential safeguards.
Looking Ahead
The immediate consequence for Faiz Enterprise is that the GST dispute will have to be reconsidered from the beginning. The department has been directed to issue a fresh notice and conduct the proceedings afresh, while properly considering the taxpayer’s response and issuing a reasoned order. The underlying tax issue therefore remains open even though the earlier proceedings have been quashed.
For India’s wider legal and administrative system, the case could become a reference point for responsible AI use in quasi-judicial proceedings. Gujarat’s existing AI policy already requires verification of AI-generated material and places responsibility on the human user. The GST case demonstrates why those safeguards matter as AI moves from experimental use into official government workflows.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
