SEBI releases draft guidelines for responsible AI use in markets; seeks public input by July

In a landmark move to regulate the growing use of Artificial Intelligence (AI) and Machine Learning (ML) in India’s financial markets, the Securities and Exchange Board of India (SEBI) has issued a draft framework proposing guiding principles for their responsible usage. The consultation paper, released on June 20, 2025, calls on stakeholders and the public to submit comments by July 11.

With AI/ML tools increasingly used by stock exchanges, brokers, and mutual funds for tasks ranging from algorithmic trading and client onboarding to surveillance and cybersecurity, SEBI’s paper highlights the need for clear governance and regulatory oversight. “AI and ML have transformative potential, but their use must be aligned with principles of fairness, accountability, and transparency,” SEBI noted.

The proposed framework rests on five core pillars: model governance, investor protection through disclosure, testing and monitoring, fairness and bias mitigation, and data privacy and cybersecurity.

Under model governance, SEBI mandates that market participants maintain in-house teams with adequate technical expertise to oversee AI models throughout their lifecycle, from development and deployment to ongoing monitoring and auditability. Firms must also set up fallback plans for critical systems and designate senior management to be accountable for AI oversight.

In terms of investor protection, SEBI recommends full disclosure when AI/ML systems directly affect clients, particularly in areas such as portfolio management, trading decisions, and financial advisory. Disclosures must include the model's purpose, limitations, data quality, and applicable fees, delivered in language that clients can understand.

The framework also outlines a robust testing regime, requiring firms to validate models in test environments before launch and monitor them continuously thereafter. Shadow testing with live data is encouraged, and firms must retain input-output logs for at least five years.

To address fairness and bias, SEBI instructs firms to use diverse data sets, proactively identify biases, and train staff to detect algorithmic discrimination. Meanwhile, in the area of data security, SEBI emphasises compliance with privacy laws and mandatory reporting of breaches or glitches.

A tiered approach has been proposed: a stricter framework for AI tools that directly impact clients, and a lighter compliance path for internal-use applications like surveillance or compliance monitoring.

SEBI has invited all stakeholders to share feedback through its online public comment portal or via email to designated officials.

With this move, SEBI becomes one of the first regulators in India to lay out an AI-specific governance roadmap for financial markets—signalling its intent to balance innovation with investor safety.