AI Regulation Tracker / Financial services

SEBI Proposes That Securities Firms Own the Liability for Their AI

In a consultation paper dated June 20, 2025, India's securities regulator floated guidelines for responsible AI and machine-learning use. The proposal puts accountability squarely on the market participant, including for AI bought from vendors. This is a consultation, not a rule, but the direction is unambiguous.

By Anthony Guerriero · The Leveraged Years Regulatory Research · Published July 21, 2026 · Last updated July 21, 2026

The short version

On June 20, 2025, the Securities and Exchange Board of India published a consultation paper titled Guidelines for Responsible Usage of AI and ML in Indian Securities Markets, with comments due by July 11, 2025. The core proposal is accountability. A market participant that uses AI or machine learning stays responsible for what it does, and SEBI would treat AI services bought from a third-party vendor as if the participant provided them itself. The paper also proposes senior-management oversight of models, independent auditing, ongoing monitoring, periodic sharing of model accuracy results with SEBI, and investor-protection disclosures. Read the status carefully: this is a consultation paper, not a binding regulation. It signals SEBI's intended direction rather than imposing duties today. For US asset managers, brokers, and AI vendors active in Indian securities markets, it is a clear preview of where accountability will sit, which is with the regulated firm, not the model or the vendor.

Regulatory Briefing

Instrument: Consultation Paper on Guidelines for Responsible Usage of AI and ML in Indian Securities Markets
Authority: Securities and Exchange Board of India (SEBI)
Jurisdiction: India; would apply to SEBI-regulated market participants such as exchanges, brokers, and mutual funds
Bindingness: Non-binding. This is a consultation paper seeking public comment. It proposes guidelines and does not create enforceable obligations on its own. Any binding rule would follow later through SEBI's normal rule-making.
What Changed: SEBI put on the table a clear accountability principle, that the regulated participant owns AI and ML outcomes including from vendors, together with oversight, audit, monitoring, and disclosure expectations.
Who This Affects: SEBI-regulated market participants and the third-party AI and ML vendors that supply them, including US firms operating in or serving Indian securities markets.
Status: Published June 20, 2025; public comments were invited until July 11, 2025. Consultation stage, pending any final guidelines.
Primary Source: SEBI consultation paper, sebi.gov.in
Last Verified: July 21, 2026

SEBI has watched AI and machine learning spread through Indian securities markets and it has decided to say something about who answers for it. On June 20, 2025, the regulator published a consultation paper called Guidelines for Responsible Usage of AI and ML in Indian Securities Markets and asked for comments by July 11, 2025. Market participants already use these models across customer support, KYC, onboarding, transaction monitoring, fraud detection, and order execution. The paper's central move is to make sure that when one of those models goes wrong, the accountability does not evaporate into the software.

The language on liability is direct. The paper states that AI and ML services provided by third-party vendors "are deemed to be provided by the market participants, who shall be responsible for ensuring compliance with all applicable laws, rules and regulations." That is the sentence to underline. Buying your AI from an outside vendor does not buy you out of responsibility for it. In SEBI's proposed framing, a bought model is treated as your model for compliance purposes, which closes the gap a firm might otherwise use to point at its vendor when a system misfires.

What is SEBI actually proposing?

The paper sets out guiding principles across the AI lifecycle rather than a single rule. It proposes that a designated senior manager with the right technical knowledge oversee model development, validation, testing, deployment, monitoring, and controls. It proposes that AI and ML systems be subjected to independent auditing by a team with no role in building them, with audit findings shared with SEBI. It proposes ongoing monitoring, since models that learn from live data can drift after deployment, and it proposes that participants share accuracy results of their AI and ML models with SEBI on a periodic basis. There is an investor-protection and disclosure strand as well, aimed at making sure customers understand when and how AI is being used on them. Underneath all of it runs the accountability principle, which is that the regulated participant, not the tool, carries the duty.

Is this binding yet?

No, and it is important to be exact about that. This is a consultation paper. SEBI is testing proposals and gathering views, not enforcing rules. Nothing in the paper imposes a duty on a market participant today, and the comment window closed on July 11, 2025. Any obligations would arrive later, if and when SEBI converts the consultation into final guidelines or a circular through its normal process, and the final text could shift in response to industry feedback. Anyone describing this as a current SEBI mandate is getting ahead of the facts. What it is, reliably, is a strong signal of where the regulator intends to land, and the accountability principle at its center is the kind of position regulators rarely walk back.

Why US professionals should read it now

Two groups have a direct stake. US asset managers, brokers, and fintechs that operate in or serve Indian securities markets should expect that, once finalized, the duty for AI outcomes will sit with the regulated Indian participant, which means your Indian operations cannot lean on a vendor to absorb compliance risk. US AI and ML vendors selling into Indian market participants should read the vendor-attribution language closely, because your clients will pass SEBI-shaped expectations back to you through contracts, audit rights, and monitoring and accuracy-reporting obligations, even though the duty formally rests with them. For financial professionals more broadly, this consultation is a clean example of the accountability pattern spreading across markets. Regulators are converging on the same answer to the question of who owns an AI decision, and the answer is the regulated firm. Building your governance around that assumption now, while the Indian rules are still a proposal, is far easier than retrofitting it after a final circular lands.

Questions professionals are asking

Is the SEBI AI paper a binding rule?

No. It is a consultation paper published June 20, 2025, with comments invited until July 11, 2025. It proposes guidelines and gathers views. It does not impose obligations on market participants today. Any binding requirement would come later through SEBI's normal rule-making, and the final text could differ from the proposal.

Who would be liable for AI outcomes under the proposal?

The regulated market participant. The paper proposes that AI and ML services provided by third-party vendors are deemed to be provided by the market participants, who shall be responsible for ensuring compliance with all applicable laws, rules and regulations. Buying AI from a vendor would not shift the compliance duty to that vendor.

What controls does SEBI propose around AI models?

Senior-management oversight of model development, validation, testing, deployment, and monitoring; independent auditing by a team not involved in building the model, with findings shared with SEBI; ongoing monitoring for model drift; periodic sharing of model accuracy results with SEBI; and investor-protection disclosures about how AI is used.

Should US firms and vendors care about it now?

Yes. US asset managers, brokers, and fintechs operating in Indian securities markets, and US AI vendors serving them, should treat the accountability principle as the likely outcome. Regulated Indian participants would carry the duty, and they will pass SEBI-shaped audit, monitoring, and accuracy-reporting expectations back to vendors through contracts.

RELATED BRIEFINGS

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any final SEBI guidelines apply to your situation with qualified Indian securities counsel.