Everyone’s been leaning on AI for legal research lately — law students, junior associates, even some senior lawyers if we’re being honest. It saves time, and honestly most of us have gotten used to just asking a tool to find a relevant precedent instead of digging through databases ourselves. But a Supreme Court judgment titles Pooja Ramesh Singh from July this year is a pretty blunt reminder of what happens when nobody actually checks what the AI hands back.
What Actually Happened
The case itself came out of an insolvency dispute, which isn’t really where you’d expect a landmark AI ruling to show up. A suspended director of Essel Infraprojects Ltd. had challenged an order passed by the National Company Law Tribunal (NCLT), Mumbai, which had admitted a Section 7 application under the Insolvency and Bankruptcy Code, 2016. The National Company Law Appellate Tribunal (NCLAT) went ahead and upheld that order when it was appealed.
When the matter finally reached the Supreme Court, something odd came up during the hearing. Several of the “precedents” the NCLT had relied on to justify its decision simply didn’t exist. Case names were fabricated, and some paragraphs were attributed to real citations that never actually contained that language. What made it worse was that the NCLAT hadn’t caught any of this either while affirming the NCLT’s order — it just went straight through without anyone checking.
What the Court Held
In Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr. (2026 INSC 668), a bench of Justice Pamidighantam Sri Narasimha and Justice Alok Aradhe didn’t hold back at all. The Court held that any decision resting on fake, non-existent, or hallucinated material — including AI-generated content passed off as precedent — amounts to no decision at all in the eyes of law. Not a minor defect that can be fixed later, but something that’s essentially void from the start, because it strikes directly at the sanctity of the adjudicatory process.
What really stood out here is how far the Court took this standard. It didn’t say the fake material had to be central to the reasoning for the decision to fall apart. Even if hallucinated content played only a small or indirect role, that’s enough to justify setting the whole decision aside. That’s a fairly strict rule, and it puts real pressure on both judges and lawyers to actually verify what they’re relying on instead of trusting a tool blindly, just because it sounds authoritative.
The Bench also went further and directed the Bar Council of India to set up a committee to look into this issue seriously — including the question of what disciplinary action should follow when a member of the bar submits fabricated material to a court, whether that happens knowingly or just carelessly.
Coming back to the actual facts, the Court set aside both the NCLT and NCLAT orders and sent the Section 7 application back to the NCLT to be decided afresh, without expressing any opinion on who’s actually right in the underlying insolvency dispute. The Court was clear that its concern here was purely about preserving the integrity of the process, not about deciding the merits of the case itself.
Why This Matters Beyond Just This Case
This case isn’t really about insolvency law at the end of the day. It’s about how much trust courts — and by extension, lawyers and students — can place in AI tools without a human actually verifying the output. The judgment doesn’t reject the use of AI in legal research altogether; it actually acknowledges that AI can be useful in assisting the process. What it draws a hard line around is human verification staying at the centre of every decision, no matter how convincing the AI output looks.
For law students especially, the practical lesson here is fairly straightforward. If you’re using an AI tool to look up case law for a moot, an assignment, or even just general research, that citation needs to be checked against an actual database like SCC Online or Indian Kanoon before you use it anywhere. It’s honestly a similar kind of mistake to how people mix up discharge and acquittal without realising they mean completely different things — except here the stakes are a lot higher, because it’s an actual court relying on something that was never real to begin with.
Conclusion
The Pooja Ramesh Singh ruling is probably going to get cited a lot over the next few months, not because insolvency law changed in any major way, but because it’s one of the clearest statements yet from the Supreme Court on where AI fits into Indian judicial process. The message really isn’t complicated — AI can help you get to an answer faster, but it can’t be the thing that decides what’s actually true, and nobody gets to skip the step of checking that anymore.
Written by Khushi Mittal
