Supreme Court Headlines Unregulated Use of AI in Judiciary | Don't Want AI to Overpower Judicial Decision Making: CJI Surya Kant
5 December 2025, 12:23 PM
The Supreme Court heard a plea seeking regulation of what the petitioner called unregulated use of AI and machine-learning tools in the judicial system. The Bench allowed withdrawal after indicating the issue is best handled administratively, not through judicial directions.
Chief Justice of India Surya Kant was clear: AI can assist, but decision-making will not depend on it. He said, "We use AI in a very conscious manner. We don't want it to overpower judicial decision making. If you have suggestions, give them on the administrative side."
Advocate Subhash Chandran argued that AI-generated material has led to errors in trial courts, including citations to Supreme Court precedents that do not exist. The CJI responded that judges are trained to verify authorities and that the system is improving: "Judges must cross check. This is already part of judicial training. With time, both the Bar and the Bench will learn."
The petitioner referenced a Supreme Court white paper on AI and guidelines issued by the Kerala High Court. The Bench reiterated its stance: the Court will not issue directions on the judicial side and invited inputs on the administrative side instead. When the Bench showed disinclination to entertain the plea, the petitioner sought withdrawal; the Court allowed it, noting the plea appeared more optics than adversarial.
What this means for legal teams
- AI is a support tool, not an arbiter. Human judgment remains final.
- Expect greater emphasis on verification: every AI-assisted output (citations, summaries, translations) must be cross-checked against authoritative sources.
- Guidance will likely mature through administrative channels (court committees, registries, training programs) rather than binding judicial orders.
- Bar and Bench training will continue to expand; firms should run parallel internal training to match court expectations.
Where AI fits in practice today
Used well, AI trims busywork and gives lawyers more time for strategy. Here are practical, defensible use cases already common in practice:
- Document review and e-discovery: Classify, de-duplicate, and surface relevant material faster in litigation, diligence, and compliance.
- Legal research: Faster retrieval of contextually relevant case law, statutes, and regulations. Established providers such as LexisNexis and Westlaw Edge now ship AI features, alongside platforms like Lexlegis.ai and CaseMine.
- Contract analysis and management: Review, compare, and flag clauses; spot risk; track obligations. Tools like SpotDraft and Kira Systems help standardize playbooks and speed redlines. LawGeex and Casetext support drafting and review workflows.
- Predictive analytics: Use historical decisions and trend data to estimate case trajectories, assess settlement ranges, and set client expectations.
- Chatbots and virtual assistants: Intake, FAQs, scheduling, and form-filling to cut administrative load and improve response times.
Practical guardrails to adopt now
- Citation discipline: Always verify with official reporters, authentic PDFs, or court websites. No AI-generated citation should enter a pleading unchecked.
- Source transparency: Track which sections were AI-assisted. Keep prompts, outputs, and validation notes in the file.
- Data hygiene: Do not paste privileged, confidential, or regulated data into tools without a vetted enterprise agreement and clear retention controls.
- Bias and scope limits: Decide in writing what AI can and cannot do in your matters (e.g., research drafts allowed; final analysis and conclusions by attorneys only).
- Tool vetting: Run security, accuracy, and explainability checks before firmwide rollout. Re-test after vendor updates.
- Human-in-the-loop: Require review sign-offs for any AI-assisted work product leaving the firm or being filed in court.
- Training cadence: Include AI use, citation verification, and confidentiality refreshers in ongoing CLE and internal programs.
If you want to contribute policy input
The Court has encouraged administrative engagement. Channel suggestions through bar associations, High Court committees, or the Supreme Court's administrative mechanisms. Focus proposals on verification standards, disclosure norms for AI use in filings, and model training for judges and registries.
Upskill without the hype
Build AI fluency that serves client outcomes: research faster, draft cleaner, verify everything. For structured learning by job role, see Complete AI Training: Courses by Job.
Your membership also unlocks:
