SC’s AI draft rules: SC lawyers’ body seeks transparency, Muslim law board flags ‘historical bias’

The Muslim law board says that AI trained on historical data could replicate prejudices based on religion, caste, gender and class biases rather than eliminating them.

Illustration: Shruti Naithani/ThePrint

New Delhi: The All India Muslim Personal Law Board (AIMPLB) has urged the Supreme Court to dilute or reconsider its proposed rules on the use of artificial intelligence in courts, saying that AI could undermine constitutional rights, harm marginalised communities and reinforce social inequalities.

The board said in a 14-page response to the Supreme Court’s AI Committee on 9 July that while it appreciates the court’s attempt to introduce safeguards such as human oversight, transparency and accountability, the draft regulations underestimate the risks posed by AI systems in the justice delivery process.

The board was established in 1973 to protect and promote the application of Shariat (Islamic law) in matters like marriage, divorce, and inheritance.

The board argued that AI models are trained on historical data that often carries existing prejudices based on religion, caste, gender and class. Consequently, algorithmic recommendations, even if described as merely “administrative”, could reinforce discriminatory patterns instead of eliminating them, the response states.

One of the board’s principal objections is that the Draft Regulations for Use of Artificial Intelligence in Courts, 2026, treat functions such as case listing, scheduling, categorisation and allocation as administrative exercises.

The board contends these decisions have significant legal consequences as delays or prioritisation in listing cases can affect liberty, bail, livelihoods and access to justice, making them unsuitable to be left to the mercy of an algorithm.

Significantly, the board said that marginalised communities like Muslims, Dalits and Adivasis may bear the brunt. The response says that Muslims and other historically disadvantaged groups have often relied on judicial scrutiny to correct executive excesses.

Introducing AI into court administration risks embedding systemic disadvantages in ways that might be difficult to detect or challenge, it said.

It adds that “the illusion of independent judgment is preserved on paper but in practice, the algorithm has already decided what the judge will see”. In case there is an error in the algorithm, the judge and the lawyer may not even discover it, the board said.

The board also pointed out that certain communities like Muslims, Dalits and Adivasis have faced disproportionate rates of arrest, prosecution, conviction and pre-trial detention owing to systemic bias at the investigative stage.

Using AI trained on judicial data drawn from these biased datasets could reflect, if not amplify, those patterns, the board said, while adding that these biases will not be erased but rather stamped with a certificate of “algorithmic neutrality”.

SCAORA’s stance

The Supreme Court Advocates-on-Record Association (SCAORA) has also sought reconsideration of some provisions of the AI draft regulations released by the Supreme Court. While welcoming the judiciary’s initiative to regulate AI, the lawyers’ body said its objections are aimed not at the technology itself but at what it described as gaps, contradictions and constitutional infirmities in the draft framework.

“Essentially, we want some transparency about what AI tools are being used. This should not be a trade secret. It’s a machine; so it’s only natural to share the information with us,” SCAORA president Devvrat told ThePrint. He added that the judiciary should disclose the AI systems already in use before the regulations come into force, arguing that lawyers and litigants cannot meaningfully evaluate tools that operate without public scrutiny.

“The present state of technology has no guardrails, so we must be cautious with it. AI is essentially a black box. We don’t know what methodology is being used to arrive at the conclusion,” said advocate-on-record Talha Abdul Rahman, who was part of the AI sub-committee which drafted SCAORA’s response.

“Artificial intelligence cannot explain itself. We are essentially being asked to trust conclusions without understanding how they were reached. That is incompatible with judicial decision-making,” Rahman said, adding that opaque or “black box” systems should have no place in high-risk court functions. “AI hallucinations and risks to human privacy” are also real concerns that come with the use of AI, he said.

Rahman said SCAORA has, therefore, recommended that all AI-generated outputs used in courts must remain subject to mandatory human oversight, with deployment introduced gradually through a precautionary, phased approach rather than wholesale adoption.

SCAORA said in its 96-page response to the regulations that it is not opposed to artificial intelligence in courts. In fact, the report welcomes the judiciary’s initiative to regulate AI.

However, its opposition is directed at what it sees as gaps, contradictions and constitutional infirmities in the draft regulations.

Among its objections, SCAORA said the regulations do not clearly identify the constitutional source of the Supreme Court’s power to frame a nationwide AI governance framework.

It argued that Article 145 is confined to Supreme Court’s own practice and procedure, and suggested that the court either invoke Article 142 for interim directions or treat the regulations as model standards for high courts to adopt under their own rule-making powers.

Firstly, SCAORA cautioned that while the draft regulations recognise AI as an assistive tool, they underestimate the risk of “automation bias”, where judges and court staff may increasingly defer to machine-generated outputs under heavy workloads.

The AOR body has also urged that AI’s role remain strictly advisory, with all outputs independently verified by humans. It has also opposed any role for AI in adjudication or administrative decisions such as case listing and prioritisation.

The association further criticised the draft’s approach to data protection, saying it relies on the Digital Personal Data Protection Act even though key obligations under the law have not yet become operational.

Instead, it has called for a court-specific data protection framework and clearer identification of who would be responsible as the data fiduciary and data processor for judicial AI systems.

SCAORA also argued that AI-generated explanations do not reflect the actual reasoning process behind their outputs. Hence, it can be used as a tool to assist judges, and not one that quietly reshapes judicial decision-making.

The lawyers’ body has also objected to the governance architecture proposed under the regulations, saying the practising Bar, the judiciary’s largest institutional stakeholder, has been given only token representation.

It has recommended making the SCAORA president, or a nominee, a permanent ex officio member of the apex AI body and involving the Bar in designing training programmes for advocates on the use of AI.

SCAORA convener Rajat Mittal, who was one of the main drafters of the SCAORA response, told ThePrint that the Supreme Court lawyers’ body has sought stronger accountability mechanisms, including mandatory external audits of high-risk AI systems, public disclosure of impact assessments whenever systems are materially modified, and removal of the requirement that advocates disclose AI use in filings. Instead, it has proposed requiring lawyers to certify that every citation and authority relied upon has been personally verified.

(Edited by Sugita Katyal)