The Seamless Link: AI's measured entry into arbitration workflows
The judiciary's engagement with AI is picking up, and India just marked a notable step. The Madras High Court has approved a controlled trial of an AI tool, "Superlaw Courts," limited to arbitration matters. The move signals interest in efficiency, without handing any part of judgment to software. Precision, guardrails, and human oversight sit at the center.
What the Madras High Court approved
On January 28, 2026, Justice N Anand Venkatesh allowed the tool's use strictly as a "record assistant." The system may organize documents, generate searchable indexes, and surface relevant excerpts on request. It is bound to the case record and barred from making inferences, evaluating credibility, or issuing any form of judicial opinion.
The pilot runs in arbitration cases involving Gammon-OJSC Mosmetrostroy JV and Chennai Metro Rail Limited. The court's message is clear: use AI to cut administrative friction, keep judgment calls with humans, and verify outputs.
Market footnote: Gammon India's numbers
While the legal development is notable, investors will parse Gammon India Ltd.'s financial position separately. The stock trades around ₹1.65 with a market cap near ₹53.59 crore, a negative book value of roughly ₹-219.17 crore, and no meaningful P/E. ROE is 11.59% and ROCE 1.11%, which do not offset broader stress. This is information for context, not investment advice.
The broader context: LegalTech growth and risk
Legal AI is a growing market, with forecasts ranging from $7.99 billion to $94.5 billion by 2030. In India, projections suggest the broader legal services market could reach ₹1,000 billion, with AI playing a larger role in research, review, and drafting. The upside is speed and consistency; the risk is error, bias, and leakage of sensitive data.
Courts have already flagged issues like fabricated citations from generative tools. That is why governance frameworks and clear operating boundaries matter. For reference, see the NIST AI Risk Management Framework (NIST AI RMF) and the UK judiciary's practical guidance on AI use (Judiciary AI Guidance).
Practical playbook: using "record assistant" AI safely
- Scope control: Limit the tool to indexing, search, and excerpting within the certified record. No legal analysis, credibility assessments, or drafting of orders.
- Data boundaries: Bind inputs to the case file; prohibit external data pulls. Disable internet access if not essential.
- Access and audit: Role-based access, encrypted storage, and immutable logs of prompts, documents accessed, and outputs retrieved.
- Validation workflow: Every excerpt or summary is verified against the source record by a human reviewer before courtroom use.
- Disclosure: Inform parties when AI is used as an administrative aid. Provide a channel to contest erroneous excerpts.
- Bias and error checks: Run test packs with known answers; track false positives, omissions, and misquotes.
- Procurement diligence: Vendor security posture, data residency, incident response, uptime SLAs, and clear IP/privilege clauses.
- Retention and deletion: Define how long processed data and logs are kept; ensure secure deletion after matter closure.
What to track in the pilot
- Time saved: Hours reduced in document organization and retrieval per matter.
- Accuracy: Precision/recall of document hits; misquote and omission rates.
- Human corrections: Percentage of AI outputs edited or rejected, and top reasons.
- User confidence: Structured feedback from judges, registrars, and counsel.
- Security: Any access anomalies, failed authentications, or data movement outside policy.
- Cost-to-benefit: Licensing and integration costs versus measurable time and error reductions.
Company actions around Gammon India
Separate from the court pilot, Hazoor Multi Projects has submitted binding offers to acquire part of Gammon Engineers and Contractors Private Limited, a Gammon India subsidiary. Punjab National Bank has also initiated an auction of bad loans that include exposure to Gammon India. These moves indicate restructuring activity and heightened lender scrutiny.
Future outlook: augmentation over substitution
AI will take on repetitive tasks-organizing records, surfacing citations, producing first-pass summaries. Lawyers and judges will focus on strategy, advocacy, and judgment. The Madras High Court's approach sets a clear template: tight scope, transparent use, and continuous verification.
If your chambers or firm is formalizing AI training for staff, you can review curated options by role here: AI courses by job. The goal is straightforward: deploy tools that reduce friction while protecting fairness, confidentiality, and the quality of justice.
Your membership also unlocks:
