The April 30, 2025 deadline for retrofitting AI disclosure into existing investment adviser and research analyst client agreements has passed. Investment advisers and research analysts using AI tools without having updated their client agreements are currently in active non-compliance with SEBI Circular SEBI/HO/MIRSD/MIRSD-PoD-1/P/CIR/2025/003. SEBI's February 2026 Master Circular for Research Analysts has since added further requirements at the point-of-delivery of individual research reports.
SEBI Circular Issued
SEBI/HO/MIRSD/MIRSD-PoD-1/P/CIR/2025/003. AI tool disclosure mandated in IA client agreements and RA terms of engagement. New client agreements: immediate. Existing clients: April 30, 2025 deadline.
Existing Client Retrofit Deadline
Deadline for updating existing IA/RA client agreements with AI disclosure language. Deadline has passed. Entities that missed it are in current non-compliance.
SEBI Master Circular for RAs
Extended AI disclosure to research reports themselves (point-of-delivery). Added AI tool change notification obligation. Required human review attestation on each AI-assisted report.
Active Compliance Obligation
January 2025 circular obligations remain in force. Every new client engagement requires disclosure. Every research report produced with AI assistance requires disclosure under February 2026 Master Circular.
Compliance Matrix: What to Disclose, Where, and When
| Disclosure Requirement | Applicable To | Document / Stage | Source |
|---|---|---|---|
| Whether AI tools are used in advisory / research process | Investment Advisers (IAs) and Research Analysts (RAs) | Client Agreement (IA) / Terms of Engagement (RA) — before advisory relationship begins | Jan 2025 Circular |
| Nature of AI use: recommendation generation vs. administrative use | IAs and RAs | Client Agreement / Terms of Engagement — at engagement stage | Jan 2025 Circular |
| Material limitations of AI tools used | IAs and RAs | Client Agreement (general limitations); Research Report (report-specific limitations) | Both |
| Responsibility-stays-with-adviser statement | IAs and RAs | Client Agreement / Terms of Engagement — explicit statement that IA/RA retains accountability regardless of AI use | Jan 2025 Circular |
| Per-report AI use disclosure | Research Analysts | Each research report produced with AI assistance — disclosure in the report itself, not only in terms of engagement | Feb 2026 Master Circular |
| Human review attestation | Research Analysts | Each AI-assisted research report — attestation that a human analyst reviewed before publication | Feb 2026 Master Circular |
| AI tool change notification to clients | Research Analysts | Client notification when material AI tool used in research changes (vendor switch, substantial model upgrade) | Feb 2026 Master Circular |
| DPDP Act consent for client data processed by AI tools | IAs and RAs processing client financial data with AI | DPDP consent notice — separate from SEBI disclosure; required before AI tool processes client data | Both + DPDP Act |
Four Liability Streams When AI-Generated Advice Goes Wrong
Client Complaint — Investor Grievance Redressal
Client who suffers loss on AI-assisted recommendation files complaint through SEBI's SCORES platform. IA/RA is accountable party. Defense requires: (a) AI tool disclosed in agreement, (b) AI output reviewed by human before delivery, (c) limitations disclosed, (d) recommendation matched client risk profile. Undisclosed AI use collapses the defense at step (a) — strongest complaint position for the client.
SEBI Inspection — Transparency Obligation Breach
SEBI inspections of IA/RA practices now review AI tool use documentation. Undisclosed AI in advisory process is a breach of IA Regulations 2013 transparency obligations. Post-April 2025, an inspection finding that agreements were not updated by the deadline is direct evidence of non-compliance. SEBI can issue show-cause, impose penalty, or suspend registration.
Mis-Selling — Research Methodology Misrepresentation
AI-generated research marketed as proprietary human analyst research — without AI disclosure — is potential mis-selling under SEBI (Research Analysts) Regulations 2014 prohibition on misleading representations about research methodology. Client paid for human analytical expertise; received AI-generated output without disclosure. Mis-selling exposure is separate from and additional to the disclosure non-compliance.
DPDP Act — Client Data Processed by AI Tool
AI tools processing client financial data (portfolio data, transaction history, financial goals) for recommendation generation are data fiduciaries under DPDP Act 2023. Separate consent obligation for AI processing of client data — not covered by standard IA client agreement. Dual exposure: SEBI disclosure non-compliance (for using undisclosed AI) and DPDP consent non-compliance (for processing client data without consent for AI use). Both regulators can act.
Retrofit Program for Investment Advisers and Research Analysts Currently in Non-Compliance
The April 30, 2025 existing-client retrofit deadline passed. Entities that did not update agreements are in current non-compliance. These five steps constitute a remediation program that should be completed before the next SEBI inspection cycle.
- Audit AI tool inventory. Document every AI tool used in any part of the investment advisory or research production process: recommendation engines, portfolio analytics AI, document summarization tools, sentiment analysis for research reports, client communication drafting AI, compliance screening AI. The disclosure obligation covers the full scope of AI in the investment process — not only tools that generate the final recommendation or report. Tools used for administrative functions (scheduling, internal communication) are outside scope, but many "productivity" tools (e.g., AI document drafting for research) are inside scope.
- Classify each tool by disclosure weight. The January 2025 circular distinguishes high-disclosure tools (AI generating recommendations or research conclusions) from lower-disclosure tools (AI used for administrative tasks). For each inventoried tool, determine: does this tool contribute to investment advice or research conclusions? If yes, it requires the full disclosure including material limitations. If no (pure administrative use), a general AI tool disclosure without specific limitations is adequate. Misclassifying a recommendation-contributing tool as administrative is a non-compliance risk.
- Draft updated agreement language for each client category. IA client agreements require disclosure in the mandatory agreement format under IA Regulations 2013. RA terms of engagement require disclosure in the terms document. Draft separate language for: IAs with individual clients, IAs with HNI/institutional clients, RAs with paid subscriber relationships, RAs with institutional research relationships. The material limitations language must be specific — generic "AI may be inaccurate" disclaimers are insufficient. Each tool's known limitations (data vintage, model domain coverage, accuracy constraints on specific asset classes) must be named.
- Implement per-report AI disclosure for research reports (February 2026 obligation). Every research report produced with AI assistance needs a disclosure section and a human review attestation. This is an operational change to research production workflow — the disclosure must be added before publication, not as a retroactive footer. Workflow design: the human analyst sign-off step must explicitly capture that they reviewed AI-contributed content; this attestation language belongs in the report's disclosure section and in the internal audit trail.
- Establish DPDP Act consent for client data processed by AI tools. Map which AI tools process client financial data. For each, confirm DPDP Act consent exists: was client consent obtained for AI processing of their financial data, or does the existing consent notice cover AI tool processing? Most consent notices issued before DPDP enforcement expected AI to be in scope implicitly — but DPDP Act 2023 requires explicit consent for specific data processing purposes. An IA/RA that built consent notices in 2023–2024 without specifically naming AI processing must update them.
Bank-Affiliated IA/RA: MRMF Intersection
Investment advisers and research analysts affiliated with banks or NBFCs face an additional layer: RBI's Model Risk Management Framework (MRMF, June 2026) applies to the bank/NBFC entity and covers all AI models in production — including AI tools used in investment advisory or research production conducted as a bank service or by a bank subsidiary.
For bank-affiliated IAs (banks providing investment advisory services under IA Regulations) and bank-subsidiary RAs (research arms of bank groups): the SEBI AI disclosure obligation covers client-facing transparency; MRMF covers internal model governance of the same AI tools. A recommendation AI used in bank wealth management must be disclosed to clients under SEBI's January 2025 circular and must be documented, validated, and monitored under MRMF. These are parallel obligations to the same AI model — satisfied with one model governance program but requiring evidence for two separate regulatory audiences (SEBI inspection, RBI examination).