SEBI’s retail algo framework, fully mandatory since 1 April 2026, requires kill switches and five-year audit logs.
NEW DELHI: The Securities and Exchange Board of India will soon issue guidelines for the responsible use of artificial intelligence in capital markets, Chairman Tuhin Kanta Pandey said on 14 November 2026. Speaking at the India International Trade Fair, he outlined three pillars: clear disclosures, accountability of regulated entities and human oversight of critical decisions. Notably, SEBI has published no rule text or effective date, so the statement signals intent.
Three pillars, one principle
Pandey said SEBI is reviewing its framework so that AI adoption does not weaken investor protection, market safety or market integrity. Press reports of the November remarks are paraphrased, and a full transcript is not yet public. Earlier, in August 2026, he said the framework would require “kill-switch and humans-in-the-loop controls along with data controls”. Furthermore, SEBI’s December 2024 proposals would keep responsibility for investor data and AI outputs with the regulated entity. That holds whether the tool is built in-house or bought from a vendor.
A rulebook built in layers
The guidelines will not start from scratch. SEBI began asking intermediaries for quarterly AI and ML reports in 2019. Then a June 2025 consultation paper proposed tiered governance. Under it, client-facing uses such as algorithmic trading, portfolio management and advisory would face full compliance. Internal uses such as compliance monitoring would face a lighter tier. Moreover, the paper proposed plain-language disclosures on purpose, risks, accuracy and fees. It also proposed independent audits reported to SEBI, bias controls and five-year retention of inputs and outputs. These remain proposals, and the final design may differ.
A dress rehearsal already running
Algorithmic trading offers the closest test. SEBI’s retail algo framework became mandatory for all stockbrokers on 1 April 2026. It makes brokers principals responsible for their algo vendors. Every strategy carries an exchange-issued Algo ID, and brokers must keep audit logs for five years. Kill switches are required, and registration applies above 10 orders per second. Furthermore, providers of opaque black-box strategies must register as research analysts. Consequently, brokers have already built authentication, tagging and logging systems that AI governance could reuse. Encouragingly, SEBI’s approach has been consistent since its first algo circular in 2012.
The regulator’s own algorithms
SEBI is deploying AI itself. Its Sudarshan supervisory platform and PolicyTwin prototype support an effort Pandey describes as making supervision increasingly predictive. In October 2026 SEBI teams won at the IOSCO TechSprint and at a global agentic regulator hackathon. Meanwhile, a May 2026 cyber incident mechanism follows the Financial Stability Board’s FIRE template. Likewise, SEBI is inventorying cryptographic systems ahead of quantum risks projected around 2032 to 2035.
Where AI could go wrong
Regulators and observers cite six risks. They include opaque black-box models, bias in training data, malicious generative AI such as deepfakes, vendor concentration, mis-selling and data-privacy breaches. Consequently, the consultation proposed that firms report their third-party AI vendors to SEBI so it can watch concentration. One question remains open. If an audited, compliant model still causes losses without human misconduct, how are investors compensated? Law students writing on the NLSBLR blog have floated algorithmic insurance or risk pooling.
Capital for deep tech
Pandey paired the AI message with a push on listings. SEBI constituted an 18-member working group on 4 November 2026 to review the Innovators Growth Platform. The platform has not drawn the traction SEBI wanted. The aim is to let deep-tech, biotechnology, semiconductor and space ventures with long gestation periods tap public markets earlier, with clear risk disclosures for investors. No deadline has been set.
What to watch
Several signals will show how the rulebook takes shape. First, watch for a circular or a revised consultation paper, and for Gazette notification of the December 2024 liability amendments. Second, the full text of the 14 November address should fix the exact wording. Likewise, the definition of a critical decision needing human sign-off will matter to brokers and asset managers. Furthermore, exchanges and industry bodies will issue technical standards once SEBI acts, as they did for retail algo rules. Finally, the working group’s report will show whether early-stage listings open up.
For Indian markets, the direction is constructive. A regulator that tests algorithms internally and has already worked through algo trading is placing human accountability at the centre of AI rules. Encouragingly, it is also widening access to capital for frontier technology firms. If the final text stays proportionate, firms could adopt AI faster and with clearer duties. Ultimately, the guidelines will show whether innovation can scale without risk scaling at the same pace.
