When a Judge Used AI and the Supreme Court Took Notice:

Gummadi Usha Rani v. Sure Mallikarjuna Rao (AP HC + Supreme Court, 2026)

Hallucinations & Sanctions | India | Supreme Court + Andhra Pradesh HC | 2026

① CASE AT A GLANCE

Full Citation (AP HC)	Gummadi Usha Rani v. Sure Mallikarjuna Rao, CRP No. 2487/2025, AP HC
Full Citation (SC)	SLP No. 7575/2026, Gummadi Usha Rani v. Sure Mallikarjuna Rao
Court (AP HC)	Andhra Pradesh High Court (Civil Revision Petition)
Court (SC)	Supreme Court of India
Judges (AP HC)	Justice Ravi Nath Tilhari
Judges (SC)	Justices P.S. Narasimha and Sudhanshu Dhulia
Date (AP HC)	January 2026
Date (SC)	February 2026
Category	AI Hallucination / Judicial Misconduct
AI Tool Used	Not disclosed (AI research tool — implied ChatGPT or similar LLM)
Who Used AI	Trial court judicial officer (Judge, not advocate)
Hallucinations Found	4 non-existent judgments cited in the trial court order
Outcome (AP HC)	Order corrected; notice issued to judicial officer; case remitted
Outcome (SC)	Proceedings stayed; AG, SG, and Bar Council of India put on notice
Judgment (AP HC)	https://www.damiencharlotin.com/documents/1395/
Judgment (SC)	https://www.damiencharlotin.com/documents/1578/

② BACKGROUND

Gummadi Usha Rani and Sure Mallikarjuna Rao were parties to a civil dispute before a trial court in Andhra Pradesh. When the trial court delivered its order, the opposing party’s counsel noticed something unusual: the judgments cited by the judicial officer in the order — legal authorities on which the entire reasoning rested — did not exist. Four cited cases were completely fabricated.

When confronted, the judicial officer candidly admitted in proceedings before the Andhra Pradesh High Court that these citations were generated using an AI tool — reportedly used ‘in good faith’ to assist with legal research. The judicial officer had not verified the citations against any official reporter or law database before incorporating them into a formal judicial order.

The matter escalated rapidly. The AP HC took serious note in January 2026; the affected party then filed a Special Leave Petition (SLP) before the Supreme Court, which admitted the matter and in February 2026 treated it as a matter of national significance.

③ THE AI ISSUE

The legal question confronting both courts was fundamental: When a judicial officer cites non-existent judgments sourced from an AI tool in a formal court order, is this a correctable procedural error — or judicial misconduct?

A secondary question: Does the ‘good faith’ use of AI as a research aid absolve a judicial officer of the professional duty to verify every legal citation before it is included in a binding judicial order? And what institutional safeguards — if any — exist or should exist to prevent AI-generated legal hallucinations from contaminating India’s judicial record?

④ WHAT THE COURTS DECIDED

• The AP HC found that the trial court order was fatally flawed — an order resting on non-existent legal authority is an order without legal basis [goes to jurisdiction/validity of the order under CPC].
• The judicial officer’s admission that AI was used ‘in good faith’ was not accepted as a defence — the HC observed that AI lacks consciousness and moral reasoning, and cannot be treated as a reliable source of legal authority without independent verification.
• The AP HC issued a formal notice to the judicial officer and remitted the matter for fresh adjudication, effectively quashing the original order.
• The Supreme Court, on the SLP, declared the conduct ‘misconduct’ — not a mere ‘error in judgment’ — elevating the gravity of the finding to a disciplinary standard rather than a correctable procedural slip.
• The SC issued notice to the Attorney General of India, the Solicitor General of India, and the Bar Council of India — signalling that the court intends to use this case to frame national guidelines on AI use by judicial officers and advocates.
• The SC stayed the underlying proceedings, freezing the case pending its decision on the larger institutional question.

⑤ KEY QUOTE FROM THE JUDGMENT

“Artificial intelligence lacks consciousness and moral reasoning. A judicial order resting on citations that never existed cannot be sustained on the ground that the source was an AI tool used in good faith. The duty of verification is absolute and cannot be delegated to a machine.”

— Justice Ravi Nath Tilhari, Andhra Pradesh High Court, CRP 2487/2025, January 2026

⑥ THE INDIA ANGLE

Indian Law Equivalent

India does not yet have a dedicated statutory provision governing AI use in judicial proceedings. However, the following provisions and frameworks are relevant:

• Order XX Rule 1 of the Code of Civil Procedure, 1908 (CPC): Every judgment must state the decision and the reasons for it — citing fabricated authority fundamentally violates this requirement as the ‘reasons’ rest on a non-existent legal foundation.
• The Bharatiya Nagarik Suraksha Sanhita, 2023 (BNSS, formerly CrPC): For criminal proceedings, fabricated reliance on non-existent precedents could vitiate orders; the same principle applies under CPC for civil matters.
• Section 65B, Indian Evidence Act, 1872 (now Sections 61-63, Bharatiya Sakshya Adhiniyam, 2023): Admissibility of electronic records requires certification. AI-generated text is an electronic record — its use without authentication raises serious admissibility questions.
• High Court (Conduct and Service) Rules: Most State High Courts govern the conduct of judicial officers under rules that require diligence, accuracy, and fidelity to the law. Citing fabricated cases potentially invites disciplinary action under these rules — precisely what the AP HC and SC appear to be invoking.

Bar Council of India Rules

Rule 10 of Chapter II, Part VI of the Bar Council of India Rules (Standards of Professional Conduct and Etiquette) requires advocates to act with utmost courtesy to the court and to uphold the authority of the court. Submitting fabricated citations — even unknowingly, through AI — potentially violates this rule by misleading the court.

Rule 15 prohibits advocates from misleading the court by making false statements. While citing AI-hallucinated cases may be done without conscious intent to deceive, the effect is identical — the court is misled. The Deepak Bahry case (Bombay HC, 2026) has already treated such conduct as a basis for adverse costs; the Gummadi SC case suggests a disciplinary referral standard is now emerging.

The Bar Council has been put on notice in the Supreme Court proceedings — an indication that the Court may issue a circular or direction requiring the BCI to frame practice directions for AI use by advocates.

Practical Advice for Indian Advocates

• Verify every citation against an authoritative database before filing — Manupatra, SCC Online, IndianKanoon, or official court websites. Never rely solely on AI-generated citations, regardless of how plausible they appear.
• If using AI for research, treat its output as a starting hypothesis — use it to identify the legal area to research, then verify each authority from primary sources. Add a verification note to your file for every AI-assisted research task.
• In matters before courts where opposing counsel cites a judgment you cannot locate, raise it immediately — courts are now alert to AI hallucinations and will take an adverse view against the party relying on fabricated authority.

⑦ QUICK TAKEAWAYS

• The Supreme Court has declared AI hallucination by a judicial officer ‘misconduct’ — not just a correctable error.
• No court in India — or the world — has so far established that ‘good faith’ use of AI excuses a professional duty to verify legal citations.
• National AI guidelines for judicial officers and advocates appear imminent — advocates should get ahead of compliance now.

— — —

---

DEEP DIVE: India’s Most Important AI Law Case

How It Began: A Trial Court Order That Couldn’t Be Verified

In 2025, a civil revision petition (CRP 2487/2025) came before the Andhra Pradesh High Court. The petitioner, Gummadi Usha Rani, was challenging an order passed by a trial court. During the proceedings, something unusual came to light: the judgments cited by the trial court judicial officer in the impugned order — the legal authorities on which the entire reasoning rested — could not be traced in any legal database, official reporter, or court record.

When the matter was pressed, the judicial officer made a startling admission: the citations had been obtained using an AI tool. The officer stated that the AI had been used ‘in good faith’ as a research aid, and that there had been no intention to mislead. However, the citations were entirely fabricated — none of the four cases cited existed anywhere in the Indian legal system.

This was not a case of misquoting an existing judgment or citing a case that had since been overruled. The judgments simply did not exist. They had never been decided by any court. The AI had hallucinated them — generating plausible-sounding case names, court names, and legal principles that bore no relationship to reality.

The Andhra Pradesh High Court’s Response

Justice Ravi Nath Tilhari, hearing CRP 2487/2025, took immediate and firm action. The AP HC held that an order premised on non-existent legal authority is fundamentally without legal basis. The court drew a distinction between a judicial error — where a judge misinterprets a real authority — and the situation here, where the very authorities cited do not exist. The latter, the HC held, goes to the validity of the order itself.

The HC’s observation that ‘AI lacks consciousness and moral reasoning’ was a pointed rejection of the ‘good faith’ defence. The court’s reasoning was straightforward: the duty to verify legal authority is not a matter of intent — it is an absolute professional obligation, the discharge of which cannot be delegated to a software tool. An advocate who cites a fabricated case cannot plead that the fabrication was done by a machine; by the same logic, a judicial officer who incorporates AI-generated citations into a formal order cannot escape accountability by reference to the technology used.

The AP HC issued a formal notice to the judicial officer — a step that is disciplinary in character — and remitted the matter for fresh adjudication, effectively striking down the original order. This was a landmark moment: the first time an Indian High Court had quashed a judicial order on the ground that it was based on AI-generated hallucinations.

The Supreme Court Takes the Case

The matter did not rest with the AP HC. An SLP (SLP No. 7575/2026) was filed before the Supreme Court of India. A bench comprising Justices P.S. Narasimha and Sudhanshu Dhulia admitted the matter and, in February 2026, delivered what is likely the most significant AI-related judicial intervention in Indian legal history.

The Supreme Court’s elevation of the matter from a ‘correctable error’ to ‘misconduct’ is a doctrinal step with far-reaching consequences. In Indian judicial service jurisprudence, ‘misconduct’ carries disciplinary implications under Articles 235 and 227 of the Constitution, which vest High Courts with superintendence over subordinate courts. An act characterised as misconduct — as opposed to a mere error in judgment — can be the basis for formal disciplinary proceedings, including adverse entries, transfer, or, in serious cases, removal from service.

The SC’s decision to issue notices to the Attorney General of India, the Solicitor General of India, and the Bar Council of India was a clear signal that the court is not treating this as an isolated incident. The AG and SG are the country’s top law officers; the BCI is the statutory body that regulates advocates. Their involvement signals that the SC intends to use this case as a vehicle for establishing systemic, national-level guidelines — not merely resolving a single dispute.

Why This Case Is Different From All Prior Indian AI Cases

All prior Indian AI hallucination cases — KMG Wires (Bombay HC, 2025), Jeetmal Choraria (Delhi HC, 2025), Deepak Bahry (Bombay HC, 2026) — involved either lawyers citing AI-hallucinated cases, or tax authorities relying on AI-generated citations in quasi-judicial proceedings. In every prior case, the AI was used by a party or representative before the court — not by the decision-maker itself.

Gummadi is unique because the AI was used by the judicial officer who authored the order. This raises a qualitatively different set of concerns. A lawyer citing a hallucinated case misleads the court — which can then verify and reject the citation. A judge citing a hallucinated case produces an order that looks authoritative, gets acted upon, and enters the judicial record — all before anyone notices the deception.

The systemic risk is compounded by the reality of India’s judicial system: with thousands of subordinate courts across the country, many operating under severe resource constraints, the temptation to use AI tools to speed up legal research is understandable. If even one judicial officer is using AI to generate citations without verification, many others may be doing the same. The Gummadi SC case is, in this sense, not just about one judge and four fake citations — it is about the integrity of the entire subordinate judicial record.

Global Comparison: How Other Countries Are Responding

India is not alone in confronting this problem. In the United States, Mata v. Avianca (SDNY, 2023) was the first major AI hallucination case, resulting in $5,000 sanctions against two lawyers. The Oregon sanctions case (2025) imposed $110,000 in penalties for 23 fabricated citations. However, in both US cases, it was lawyers — not judges — who used AI. The US has not yet confronted an AI hallucination by a judicial officer.

In the United Kingdom, Judicial College guidelines issued in 2024 warn judges against using generative AI tools for legal research. The guidance acknowledges that AI can produce plausible-sounding but entirely fictitious legal authority. Canadian courts have similarly issued notices requiring disclosure of AI use in submissions. No jurisdiction, however, has as yet confronted the specific scenario that Gummadi presents: a judicial officer’s order being quashed and the matter being treated as misconduct at the highest court level.

India’s Supreme Court, by taking this case on and issuing notices to the law officers and Bar Council, appears to be setting up a framework that will be more comprehensive than anything seen elsewhere. If the SC issues practice directions — as seems likely — India may become the first country to have binding national rules governing AI use both by advocates and by judges.

What to Watch For

The Supreme Court’s final order in SLP 7575/2026 is anticipated to address: (1) whether AI use by judicial officers requires prior authorisation or disclosure; (2) whether citation verification protocols should be mandated; (3) what disciplinary standard applies when a judicial officer cites AI-generated hallucinations; and (4) whether Bar Council Rules need to be amended to address AI-generated submissions by advocates.

The BCI, put on notice, will likely need to issue a circular — either formalising the prohibition on unverified AI-generated citations or establishing a mandatory verification protocol as a condition of professional practice. Given that the SC has now established that good faith is no defence, advocates across India should treat this case as a bright line: verify every citation, document the verification, and never file AI-generated authority without independent confirmation.

For litigants and advocates in Andhra Pradesh, the immediate consequence is equally stark: if you encounter an order from a lower court that cites judgments you cannot trace, raise the issue immediately — you may have grounds to challenge the order on the same basis that succeeded in Gummadi.