Human Intelligence Must Prevail — Not Merely Be Advised
The recent decision of the Andhra Pradesh High Court in Gummadi Usha Rani v. Sure Mallikarjuna Rao raises an uncomfortable but necessary question for the Indian judicial system:
Can a judicial order survive when its reasoning is supported by authorities that never existed?
The High Court answered in the affirmative—holding that mere reference to non-existent citations generated by an Artificial Intelligence tool would not vitiate a judicial order, provided the legal principle applied is otherwise correct.
While the Court’s caution against uncritical use of AI tools is laudable, the conclusion it ultimately reaches is jurisprudentially troubling and institutionally risky.
AI Is Not the Villain — Institutional Leniency Is
There is no dispute that the trial court relied upon case laws later found to be AI-generated hallucinations, untraceable in any authoritative legal database. The judicial officer candidly admitted the error and assured greater caution in the future.
The High Court correctly observed that AI:
- lacks judicial reasoning,
- can generate persuasive but legally incorrect content,
- requires strict human verification.
Yet, having acknowledged the danger, the Court effectively neutralised its consequences.
That is the core concern.
Why “Correct Principle” Is Not a Safe Harbour
The Court reasoned that since settled legal principles were applied, the order need not be interfered with despite reliance on fictitious precedents.
This approach confuses correctness of outcome with legitimacy of process.
Judicial authority flows not merely from arriving at an acceptable result, but from:
- traceable reasoning,
- verifiable authorities,
- and transparent application of binding precedent.
A judgment is not validated simply because it reaches a defensible conclusion.
Faulty reasoning cannot be cured retrospectively by a correct result.
Precedents Are Not Decorative Footnotes
Case citations serve constitutional functions:
- enabling appellate scrutiny,
- ensuring consistency and predictability,
- providing transparency to litigants,
- and sustaining public confidence in courts.
When a judgment cites cases that do not exist, these functions collapse.
Lawyers and litigants are left chasing ghost precedents, misled by the judicial record itself.
That harm is neither theoretical nor trivial.
A Slippery Slope in the Age of AI
Today, the Court holds that fake citations alone do not vitiate an order.
Tomorrow, this logic may excuse:
- partial verification,
- casual AI-assisted drafting,
- routine reliance on unverified machine outputs.
What begins as an exception risks becoming institutional tolerance.
The real danger is not that AI will replace judges—but that verification itself may quietly be outsourced.
A Missed Opportunity
The Court itself referred to foreign and Indian precedents warning against fabricated authorities and stressing professional responsibility. Yet it stopped short of enforcing a meaningful corrective.
A more balanced course was available:
- recalling or correcting the order,
- expunging fictitious citations,
- laying down mandatory verification standards.
Such an approach would have preserved the result without legitimising flawed reasoning.
“No Prejudice” Is a Misleading Comfort
The assertion that no prejudice was caused because the Advocate Commissioner’s report is only evidentiary misses the point.
Prejudice is not only outcome-based.
There is systemic prejudice when:
- false law enters the judicial record,
- judicial time is wasted,
- and confidence in adjudication erodes.
Judicial credibility depends on precision, not benevolent intentions.
Human Intelligence Must Prevail — In Practice
The Court concluded that human intelligence must prevail over artificial intelligence. That principle is sound.
But human intelligence prevails not by merely warning against AI misuse,
it prevails by refusing to validate its errors.
If courts excuse AI hallucinations today, controlling them tomorrow will be far more difficult.
The High Court was right to caution against the dangers of Artificial Intelligence in judicial work. However, by holding that reliance on fictitious authorities does not vitiate an order, it risks normalising error in the age of automation.
In a constitutional system governed by the rule of law,
there can be no such thing as harmless fiction in a judicial order.
AI may assist the judiciary—but judicial discipline cannot be diluted, delegated, or forgiven.
Read Judgement Here
