What AI liability does SEBI's Intermediaries Amendment Regulations 2025 create for securities market intermediaries?

SEBI (Intermediaries) (Amendment) Regulations 2025 (Gazette Notification, February 6, 2025, in force February 10, 2025) inserted a new chapter 'Usage of Artificial Intelligence' into the principal SEBI (Intermediaries) Regulations 2008. The amendment applies to all SEBI-regulated intermediaries — 9,000+ registered entities including stock brokers, depository participants, investment advisers, portfolio managers, research analysts, mutual fund AMCs, collective investment schemes, and market infrastructure institutions (NSE, BSE, NSDL, CDSL, clearing corporations). The chapter creates three direct AI liability obligations: (1) Sole accountability for AI outputs: every intermediary is 'solely responsible' for all outputs generated by AI/ML tools it uses, whether the tool was built internally or procured from a third-party vendor. An intermediary cannot shift liability to an AI vendor even through contractual indemnification — SEBI regulation overrides vendor T&Cs; (2) Privacy, security and integrity of investor data: every intermediary is solely responsible for the privacy, security and integrity of investors' and stakeholders' data when processed by AI tools. An AI vendor data breach that exposes investor data is the intermediary's regulatory liability, not the vendor's; (3) Compliance regardless of AI vendor T&Cs: the intermediary is responsible for compliance with all applicable laws when using AI tools, regardless of what the AI vendor's terms of service say. An AI tool's T&C clause disclaiming liability for regulatory breaches has zero force against SEBI. This is India's first direct statutory AI liability rule for securities market intermediaries — prior SEBI circulars on AI (the 2019 ML reporting circulars for stock brokers and AMCs) required reporting of AI use but did not assign liability for AI outputs.

How does SEBI's 2025 AI chapter define AI/ML tools and what is its scope?

SEBI's Intermediaries Amendment 2025 defines 'AI and ML tools' broadly: any application, software, or system used to facilitate investing, trading, distributing investment strategies, carrying compliance, management, or business activities. This definition is deliberately broad and technology-agnostic — it covers: (1) Trade execution AI: algorithmic trading systems, smart order routing AI, and execution optimization models are clearly within scope; (2) Research and recommendation AI: AI-generated research reports, AI-driven stock screening, robo-advisory recommendation engines, and AI-powered portfolio construction tools; (3) Compliance AI: AML transaction monitoring systems, fraud detection models, KYC verification AI, and regulatory reporting automation are AI tools used to 'carry compliance activities'; (4) Customer-facing AI: chatbots, virtual assistants, and AI customer service tools used for investor queries and account management; (5) Back-office AI: AI used in settlement processing, margin calculation, corporate action processing, and reconciliation. The broad definition means an AMC using AI for NAV calculation, a broker using AI for margin computation, and a depository participant using AI for account statement generation are all covered. The companion amendments — SEBI (Securities Contracts) (Stock Exchanges and Clearing Corporations) (Amendment) Regulations 2025 and SEBI (Depositories and Participants) (Amendment) Regulations 2025 — extended the same AI chapter to market infrastructure institutions, leaving no SEBI-regulated entity outside the AI accountability framework.

What AI audit trail and documentation does SEBI's 2025 AI chapter require?

SEBI's Intermediaries Amendment 2025's 'solely responsible' accountability clause implies an audit trail obligation that the regulation does not fully define but that SEBI inspection practice makes practically mandatory: if an intermediary is solely responsible for AI outputs, it must be able to demonstrate — to SEBI, to an investor grievance forum, or to the Securities Appellate Tribunal — that the AI output was appropriate, accurate, and compliant. This creates an implicit documentation requirement across five dimensions: (1) AI tool inventory: every AI tool in use (internally built or third-party procured) must be known and documented — an intermediary that cannot tell SEBI what AI tools it uses cannot satisfy the accountability obligation; (2) AI decision audit trail: for AI outputs that affect investor interests (a research recommendation, a portfolio rebalancing decision, a margin call, a suspicious transaction report), the AI's decision must be logged with sufficient detail to reconstruct what the AI decided, based on what inputs, at what time, and with what output; (3) AI vendor assessment: before deploying a third-party AI tool, the intermediary must assess the tool for compliance with SEBI requirements — data handling, output accuracy, bias risk, security. Post-deployment assessment records constitute evidence of due diligence; (4) Incident documentation: when an AI tool produces a non-compliant output (incorrect research recommendation, incorrect margin calculation, misclassified transaction), the incident must be documented, investigated, and remediated — with evidence that the root cause was addressed; (5) Investor grievance AI attribution: when an investor complaint arises from an AI-influenced decision (AI recommended a stock that declined, AI flagged an account for review incorrectly), the intermediary's grievance response must address the AI's role — which SEBI's grievance redress framework now effectively requires intermediaries to be able to explain.

How does SEBI's 2025 AI amendment affect intermediaries using third-party AI from fintech vendors?

The majority of India's 9,000+ SEBI-registered intermediaries are not large institutions with internal AI teams. Sub-brokers, registered investment advisers (RIAs), small AMCs, and category-I and II AIFs routinely use third-party AI from fintech vendors — AI-powered research platforms, robo-advisory engines, AML screening tools, and digital onboarding AI. SEBI's Intermediaries Amendment 2025 creates a compliance asymmetry: small intermediaries now bear the same sole accountability for AI outputs as large institutions, but with far less internal capacity to govern third-party AI. Four specific governance obligations arise for intermediaries using third-party AI: (1) Vendor due diligence documentation: the intermediary must conduct and document pre-deployment due diligence on every AI tool — accuracy testing, security assessment, data handling review, compliance mapping to SEBI regulations. This due diligence cannot be outsourced to the vendor's self-certification; (2) Data processing agreements: investor data processed by third-party AI must be governed by a data processing agreement that gives the intermediary control over data use, retention, and security — and that makes the vendor contractually obligated to support the intermediary's compliance obligations (not merely disclaim liability); (3) Change management notification: when a fintech vendor updates its AI model, the intermediary's SEBI accountability requires it to assess the change's impact on compliance. Without contractual change notification rights, the intermediary cannot manage this; (4) Incident accountability: when a third-party AI vendor's system causes an investor harm (an incorrect research signal, a false AML alert that freezes an investor account), SEBI's sole accountability rule means the intermediary faces regulatory action, not the vendor. Indemnification rights in vendor contracts are between private parties — they do not shift the SEBI regulatory liability.

AI Governance · SEBI · Intermediaries · AI Liability · Securities Market · India BFSI

SEBI Intermediaries Amendment Regulations 2025 AI Liability Governance: Sole Accountability for AI Tool Outputs, Investor Data Security, Third-Party AI Vendor No Safe Harbour and India's First AI Liability Rule for Securities Market

SEBI (Intermediaries) (Amendment) Regulations 2025 (February 6, 2025) inserted India's first direct AI liability chapter into securities market regulation — making every SEBI intermediary solely responsible for AI tool outputs, investor data security under AI, and regulatory compliance regardless of AI vendor T&Cs. The amendment applies to 9,000+ SEBI-registered entities from NSE, BSE, NSDL, CDSL down to individual registered investment advisers. Zero prior SEBI circular carried direct AI liability. Zero global competitor coverage.

Zytra Research Team · February 2025 · 11 min read

📖 Part of the India BFSI AI Governance Complete Guide 2026 — RBI, SEBI, IRDAI, NPCI & DPDP in one place.

India's securities market employs AI pervasively — algorithmic trading at brokers, AI portfolio construction at AMCs, AI-generated research at research analysts, robo-advisory at investment advisers, AI-powered KYC and AML at depository participants. Prior to February 2025, no single SEBI regulation assigned direct legal accountability for AI outputs to the intermediary using them. SEBI's 2019 AI/ML reporting circulars for stock brokers and AMCs required quarterly disclosure of AI use — but not liability for AI consequences.

SEBI's (Intermediaries) (Amendment) Regulations 2025 ended this. The new AI chapter — inserted February 6, 2025, in force February 10, 2025 — makes sole accountability for AI outputs the statutory position for every SEBI-regulated intermediary. India's 9,000+ intermediaries now operate under a regulatory framework where "the AI vendor was responsible" is not a defence.

Sole Accountability: What It Actually Means in Practice

Sole accountability for AI outputs means the intermediary bears SEBI regulatory liability for what the AI produces, regardless of the contractual relationship with the AI vendor. An AMC whose AI portfolio construction tool makes systematic factor errors that harm investor returns cannot point to the robo-advisory vendor's limitation of liability clause. A broker whose AI-generated research recommendation proves materially incorrect cannot argue the AI platform's T&Cs disclaim accuracy. SEBI's regulation is a public law obligation that overrides private contract terms.

The practical implication: intermediaries must govern third-party AI as if it were internally built. Due diligence before deployment, ongoing monitoring, documented change assessment, and incident investigation are now SEBI compliance obligations — not just good operational practice.

The Small Intermediary Problem

India has thousands of registered investment advisers (RIAs), small category-I and II AIFs, and regional sub-brokers. Many operate without internal compliance teams of the scale needed to govern AI tools with the rigor that sole accountability demands. SEBI's AI chapter applies equally to all — creating a compliance burden asymmetry that SEBI has not yet addressed with proportionate guidance for smaller intermediaries.

SEBI Intermediaries AI Governance Checklist

☐ AI tool inventory: all AI/ML tools in use documented — internally built and third-party; function, vendor, date deployed, data accessed, SEBI compliance mapping
☐ Pre-deployment vendor due diligence: accuracy assessment, security review, data handling audit, SEBI compliance mapping — documented for every third-party AI tool
☐ Data processing agreements: investor data AI processing governed by DPA with control, retention, security, and compliance support obligations; not merely vendor disclaimer
☐ AI decision audit trail: AI outputs affecting investor interests logged with inputs, outputs, timestamp; reconstructable for SEBI inspection and investor grievance
☐ Change management protocol: AI tool change notification rights in vendor contracts; impact assessment on SEBI compliance for material changes; board or compliance committee approval before go-live
☐ Incident response: non-compliant AI output investigation protocol; root cause documentation; investor remediation process; SEBI notification assessment
☐ Investor grievance AI attribution: AI role in investor-affecting decisions explainable; grievance response addresses AI component; evidence on file
☐ SEBI examination readiness: AI tool inventory, vendor due diligence records, DPA documentation, AI decision audit trail — packaged for SEBI inspection