Indian courts sanctioned the use of AI-generated fake case law twice in 2026: once in a civil trial court order relying on four non-existent Supreme Court judgments, and again in an insolvency matter where the NCLT and NCLAT both relied on fabricated precedents. The Supreme Court has called this "misconduct" and adopted a zero-tolerance approach, and the accountability falls on the filing advocate, not the AI tool. For a platform built around document-verified advocates, this is a directly relevant development: verification is now an explicit professional duty, not just a best practice.
The trial court order: four judgments that didn't exist
In August 2025, a junior civil judge in a trial court in Vijayawada dismissed a party's objections to an advocate commissioner's report in a property dispute, citing four Supreme Court judgments in support. Those judgments did not exist. On 27 February 2026, a Supreme Court bench of Justices P.S. Narasimha and Alok Aradhe took suo motu cognizance of the order, holding that "a decision based on such non-existent and fake alleged judgments is not an error in the decision-making process. It would be a misconduct and legal consequence shall follow." The Court sought responses from the Attorney General, the Solicitor General and the Bar Council of India, and appointed a senior advocate as amicus curiae to assist on the wider issue.
The NCLT/NCLAT case: fabricated precedents in an insolvency dispute
On 2 July 2026, the same underlying concern surfaced again, this time at a higher stake. In Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. (2026 INSC 668), an insolvency application under Section 7 of the Insolvency and Bankruptcy Code had been admitted by the NCLT and upheld by the NCLAT. The Supreme Court found that both tribunals had relied on judgments that did not actually exist, some entirely invented, others real cases with fabricated paragraphs, bearing the hallmarks of generative-AI output. The Court set aside both orders, holding that a decision tainted by such fabricated material is "no decision in the eyes of the law," and directed the Bar Council of India to constitute a committee to frame guiding principles on AI use in litigation.
Who is actually accountable: the advocate, not "the AI"
A consistent thread across both rulings is that responsibility is pinned on the professional who filed the citation, not on the AI tool used to generate it and not on the client. Citing fake or unverified AI-generated precedents is treated as professional misconduct by the advocate concerned. This matters because it forecloses a defence some advocates might otherwise raise, that the AI tool was at fault, or that the error was a technical glitch rather than a professional lapse. Courts have made clear that the duty to verify case law before relying on it sits squarely with the advocate presenting it.
Are judges also restricted from using AI?
Separately from the advocate-facing rulings above, some High Courts have moved to restrict judicial use of AI tools directly. The Punjab and Haryana High Court has instructed judicial officers in Punjab, Haryana and Chandigarh not to use AI tools such as ChatGPT, Gemini, Copilot or Meta's tools to draft judgments or conduct legal research, stating that judicial outcomes must remain the product of human reasoning and professional expertise, not machine-generated text.
A verification checklist before filing any AI-suggested citation
Whether AI tools were used for early-stage research, drafting help, or brief summarisation, the practical takeaway from these rulings is the same: never file a citation without independently confirming it exists and says what it's being cited for. A workable pre-filing checklist:
- Cross-check against an official law report. Confirm the citation appears in SCC, AIR, or another established law report, not just in the AI tool's output.
- Verify against the court's own database. Search the Supreme Court's or the relevant High Court's own e-SCR or judgment database for the case by party name and date, not just by the citation number the AI produced.
- Read the actual paragraph being relied on. As the July 2026 ruling showed, some fabricated material takes the form of real cases with invented paragraphs inserted, so confirming a case exists is not sufficient; confirm the specific quoted or paraphrased proposition actually appears in it.
- Treat unfamiliar citations with extra scrutiny. A citation that reads plausibly but that you cannot independently locate through at least one authoritative source should not be filed, regardless of how confidently it was generated.
- Keep a verification trail. Note where and how each citation was confirmed before filing, particularly for matters where AI-assisted drafting tools were used at any stage.
This is directly relevant to the kind of verification-first practice LawyerDhundo asks of the advocates listed on this platform: every advocate here has had their enrolment and identity documents checked before being listed, and the same independent-verification discipline that this site applies to advocate credentials is exactly what these 2026 rulings now expect advocates to apply to their own AI-assisted research before it reaches a court. Delayed or careless filing has separately drawn a harder line from courts recently; see our guide on why "still researching" no longer excuses a delayed filing for a related shift toward higher due-diligence expectations of advocates.
Disclaimer: This article is for general information only and is not legal advice. Court directions on AI use in litigation are an evolving area; consult a qualified advocate or the Bar Council of India's published guidance for the current position.
Frequently asked questions
Do Indian courts treat AI-generated fake case citations as a serious problem?
Yes. In 2026, the Supreme Court flagged this twice, once taking suo motu cognizance of a trial court order that relied on four non-existent judgments, and again setting aside NCLT and NCLAT orders that relied on fabricated and doctored AI-generated precedents. The Court has described a 'zero-tolerance' approach and directed the Bar Council of India to frame guiding principles.
Who is held responsible when an AI tool generates a fake case citation that gets filed?
The filing advocate, not the AI tool or the client. The Supreme Court has held that citing fake, unverified AI-generated precedents amounts to professional misconduct on the part of the advocate who relied on and filed them, since it is the advocate's professional duty to verify the authenticity of any case law before placing it before a court.
What happened in the Andhra Pradesh trial court case?
In August 2025, a trial court in Vijayawada dismissed a party's objections in a property dispute while relying on four Supreme Court judgments that do not exist. On 27 February 2026, a Supreme Court bench of Justices P.S. Narasimha and Alok Aradhe took suo motu cognizance of the order, sought responses from the Attorney General, Solicitor General and the Bar Council of India, and appointed a senior advocate as amicus curiae.
What happened in the NCLT/NCLAT insolvency case?
On 2 July 2026, the Supreme Court set aside orders passed by the NCLT and NCLAT in an insolvency matter after finding both tribunals had relied on judgments that did not exist, some entirely invented and others real cases with fabricated paragraphs, bearing the hallmarks of generative-AI output. The Court held that a decision based on such material is not a decision at all in the eyes of the law.
Are Indian judges allowed to use AI tools to draft judgments?
Some High Courts have moved to restrict this directly. The Punjab and Haryana High Court, for instance, has instructed judicial officers not to use AI tools such as ChatGPT, Gemini or Copilot for drafting judgments or conducting legal research, stating that judicial reasoning should remain the product of human deliberation.