SC Proposes Complete Ban On AI-Driven Adjudication

The Supreme Court has released a comprehensive draft framework governing the use of artificial intelligence (AI) in the Indian judiciary, proposing strict restrictions that ensure AI remains purely assistive and never replaces human judicial decision-making.

Titled the “Regulations for Use of Artificial Intelligence in Courts, 2026”, the draft has been prepared under the Supreme Court’s AI Committee chaired by Justice PS Narasimha, and includes members from the higher judiciary such as Justices Sanjeev Sachdeva, Raja Vijayraghavan V, Anoop Chitkara, and Suraj Govindaraj. 

The top court has invited public and stakeholder feedback by June 20.

AI Cannot Decide Cases Or Influence Judgments

At the core of the proposed framework is a complete prohibition on AI-driven adjudication. The draft clearly states that no AI system can be used to decide judicial outcomes, including judgments, sentencing, or findings of fact or law.

It also bans algorithmic decision-making in courts, ensuring that the final authority in all matters remains exclusively with human judges.

“Every AI system shall function solely in an assistive capacity and shall not supplant or compromise the independent exercise of judicial authority by a duly appointed judicial officer,” the draft said.

Besides, the Committee has proposed the following restrictions: 

• Personal data restriction: Personal data cannot be used to train or test AI systems without prior approval and legal compliance.

• No AI-only judgments: AI cannot independently decide any judicial outcome; only human judges can give final decisions.

• Adjudication & sentencing control: AI cannot adjudicate or sentence; its output is only advisory and must be reviewed by a human judge.

• Ban on risk scoring: AI cannot be used to predict bail outcomes, flight risk, recidivism, or credibility of any person.

• No opaque AI systems: Black-box or unexplainable AI cannot be used in cases affecting rights or personal liberty.

• No behavioural profiling: AI cannot predict or infer future behaviour of accused persons, witnesses, or parties.

• No AI surveillance in courts: AI cannot monitor judges, lawyers, litigants, or court participants unless legally authorised.

• Mandatory disclosure of AI evidence: Any AI-generated material must be clearly disclosed and cannot be treated as independent evidence.

• Protection of judicial independence: AI must not interfere with judicial deliberations or compromise independent decision-making.

The draft also disallows any opaque or “black box” AI systems in judicial processes, particularly where fundamental rights or personal liberty may be affected.

For context, black box refers to an AI system, usually based on deep learning, whose internal decision-making process is not transparent, understandable, or explainable through clear rules or factors accessible to the user or affected person.

Where AI Can Be Used

The draft permits the use of AI in several non-adjudicatory and administrative functions: 

• Case management support: AI can help identify defects in new filings, prepare cause lists, schedule hearings, and prioritise dockets.
• Court proceedings support: It can transcribe hearings automatically, but accuracy must be checked and certified by a designated officer.
• Translation work: AI can translate legal documents, but all translations must be verified by humans.
• Legal research assistance: It can help with case law search, precedent retrieval, citation checks, and summarising documents.
• Administrative tasks: Includes filing support, defect checking, record management, and allocating judicial resources.
• Litigant support tools: Chatbots can guide users on court procedures but must be supervised by humans.
• Accessibility features: AI can support text-to-speech, speech-to-text, Braille conversion, and visual assistance.
• Fraud detection tools: Can help verify document authenticity, but outputs must be reviewed before action.
• Judgment publication support: AI can anonymise judgments and court records for public release.
• Judicial analytics: Can analyse performance, backlog, and court efficiency.
• Document automation: Can generate formats, notices, summons, and administrative documents.

The framework also sets strong safeguards and accountability rules. Judges must always supervise AI use, and they remain fully responsible for all decisions. AI can only assist and cannot give final judgments. Regular audits, monitoring, and compliance with the Digital Personal Data Protection Act, 2023 are required. Courts must also clearly disclose when AI tools are used.

A permanent Apex Body will oversee the system, including judges, technical experts, government representatives, and academics. It will set standards, approve AI tools, monitor performance, and publish yearly reports. It will be supported by committees and AI units in High Courts.

This comes at a time when the usage of AI is picking up pace across industries. In the legal industry, it is significantly reducing time spent on repetitive tasks such as document review, complaint handling, contract verification, and legal research, enabling lawyers to focus more on strategy and client advisory, with startups like Lawyered, Lucio, LawRato, LegalKart, and MyAdvo leveraging the technology.

In a separate development, the consumer affairs ministry said it is working on AI-based machine-readable “SMART” standards to reduce compliance burden on industry.

Speaking at a FICCI event, consumer affairs secretary Nidhi Khare said the department is exploring how AI can modernise India’s standards ecosystem, while also managing its disruptive impact.

“We also understand that the emerging technologies which are coming up in a big way, especially AI, may be creating a lot of disruption, and it could be creating more challenges. But we need to understand how to use it to our advantage,” she said, according to PTI.

Machine-readable standards convert rules into digital formats that systems can automatically process for compliance checks, while SMART (Standard Machine Accessible, Readable and Transferable) formats are dynamic and integrated into industry systems.

The post SC Proposes Complete Ban On AI-Driven Adjudication appeared first on Inc42 Media.