SEBI’S Approach to Artificial Intelligence and Machine Learning: Exploring SEBI’S AI/ML Consultation Paper

The Securities and Exchange Board of India (“SEBI”) has released a consultation paper dated June 20, 2025, titled “Guidelines for Responsible Usage of AI/ML in Indian Securities Markets” (“Paper”), proposing a regulatory framework for the responsible usage of Artificial Intelligence (“AI”) and Machine Learning (“ML”) tools in the Indian securities markets. While AI/ML has the potential to enhance productivity, efficiency, and outcomes, it also introduces risks that could impact market integrity and investor interests. The Paper represents SEBI’s proactive and forward-looking approach to addressing the ethical, legal, and operational challenges associated with emerging technologies associated with AI/ML.

Regulatory Approach

The Paper makes reference to the September 2021 report issued by the International Organization of Securities Commissions (IOSCO) on the use of AI (“Report”). Drawing from the principles outlined in the Report, the Paper suggests framing guidelines on model governance, investor protection and disclosure, testing, fairness and bias, data privacy, and cyber security.

Following six key measures were proposed in the Report:

• Regulators should ensure that firms have designated senior management responsible for overseeing all aspects of AI/ML usage. This should include a documented internal governance framework with clear lines of accountability.
• Firms must continuously test and monitor AI/ML algorithms to validate performance and ensure stability.
• Firms should have sufficient skills, expertise, and experience in-house to develop, deploy, monitor, and control the AI/ML tools they use.
• Firms must manage their reliance on third-party AI/ML service providers, including ongoing monitoring and oversight of performance.
• Firms should disclose meaningful information to customers and clients about the use of AI/ML, especially when it impacts client outcomes. Regulators should also require firms to furnish necessary information to ensure oversight.
• Firms must have appropriate controls to ensure high-quality, unbiased, and diverse datasets that support the effective application of AI/ML.

Usage Of AI/ML in the Indian Securities Market

AI/ML technologies are already being employed in various capacities across the Indian securities ecosystem:

• Exchanges are using AI/ML for market surveillance, cybersecurity, chatbot-based member support, automated compliance functions, social media analytics, and pattern recognition.
• Brokers are applying AI/ML tools for KYC/document verification, product recommendations, chatbots, digital account opening, transaction monitoring, surveillance, Anti-Money Laundering (AML), order execution, and mutual fund selection.
•  Mutual Funds are leveraging AI/ML for customer service including deploying chatbots, surveillance, cybersecurity, and customer segmentation.

Recommendations of the Working Group

The Paper emphasizes the importance of firms developing internal capabilities to manage the AI/ML lifecycle, including performance monitoring, model security, interpretability, and ethical deployment. The working group’s recommendations aim to ensure robust governance, risk mitigation, and responsible innovation through continuous oversight and ‘human-in-the-loop’ decision-making processes.

The guidelines are built on five core principles :

Model Governance

Market participants deploying AI/ML models must:

• maintain internal teams with adequate technical skills to oversee model development and performance throughout the lifecycle;
• implement risk control measures and robust governance, especially under stressed market conditions;
• establish procedures for exception and error handling;
• appoint senior management personnel with relevant technical knowledge to be responsible for oversight;
• carefully manage relationships with AI/ML vendors and third-party providers; and
• ensure compliance with applicable legal and regulatory frameworks.

Investor Protection and Disclosure

When AI/ML models directly affect clients or investors, firms should:

• clearly disclose the purpose, features, limitations, accuracy, and potential risks associated with such models;
• communicate information in simple language, including the quality and completeness of the data used such that the customers/clients are able to make informed decisions; and
• specify any applicable fees and maintain transparency to foster trust and accountability.

Testing Framework

Firms should:

• continuously test and validate AI/ML outputs and model performance;
• maintain thorough documentation, storing all input/output data for at least 5 (five) years;
• move beyond conventional testing methods for traditional algorithms and adopt enhanced monitoring protocols tailored to the evolving nature of AI/ML.

Fairness and Bias

To ensure ethical outcomes:

• AI/ML systems must not discriminate against or favour particular groups of clients;
• firms should implement mechanisms to identify and mitigate bias within datasets; and
• data used should be of high quality, sufficiently diverse, and representative.

Data Privacy and Cyber Security

As AI/ML systems heavily depend on data, firms must:

• have comprehensive policies for data privacy, cybersecurity, and protection of personal investor information;
• ensure compliance with applicable laws relating to data processing, breach reporting, and cyber risk mitigation; and
• promptly notify SEBI and other relevant authorities of any data breaches, system glitches, or security lapses.

For the purpose of managing the risk arising from usage of AI/ML models, possible control measures are mentioned in the Annexure B of Paper. The possible control measures for the given risk are as follows:

• Malicious usage leading to market manipulation and/or misinformation:

By watermarking and provenance tracking, suspicious activity reporting and public awareness campaigns, the risk of price manipulation and market instability by creating fraudulent financial statement, misleading news articles, or deepfake content can be reduced.

• Concentration Risk:

Any dominant AI system/model’s provider in the financial market are subject to enhanced monitoring such as more frequent reporting of performance results and audit filing. Market participants are encouraged to used multiple suppliers and required to report the names to their providers so that the regulator could monitor any build-up of concentration. This ensures that the market participant does not have reliance on limited number of Gen AI providers that could lead to systemic risks in times of failure or impairment.

• Hearding and Collusive Behaviour:

Market Participant should use varied AI architectures and proprietary databases and stock exchanges monitors the potential herding behaviour arising from the similar AI-driven strategies to prevent the potential impact on financial markets due to widespread use of common models and databases. 

• Lack of explainability:

Market participants should maintain detailed AI process documentation and use such interpretable AI models or explainability tools which can explain working of AI model so that the Gen AI models are not difficult to comprehend and not impede supervision and regulatory oversight.

• Model failure / runaway AI behaviour:

Extreme scenarios are simulated to do stress testing to assess the AI performance and to prevent the over reliance on AI systems, participants can keep human involved in decision making so that the flaws in the Gen AI system could not spread across market, leading to financial stability. 

• Lack of Accountability and Regulatory Non-Compliance:

Testing of the AI systems in controlled environments is done to ensure that they do not result in regulatory breaches to prevent compliance lapses, regulatory infractions, and investors losses particularly if their outputs are not effectively monitored. Human-in-the-loop mechanism can also be implemented to prevent over reliance on AI systems.

SEBI’s Paper is a significant and timely initiative aimed at building a responsible AI/ML regulatory framework in the Indian securities market. As the use of AI/ML becomes increasingly prevalent in financial systems, it is crucial to balance innovation with investor protection, transparency, and accountability. This consultation process invites stakeholder feedback and is expected to shape the final regulatory framework, aligning India’s capital markets with global best practices in AI governance.