What AI/ML reporting obligations do SEBI's 2019 Mutual Fund AI Circular create for AMCs?

SEBI Circular SEBI/HO/IMD/DF5/CIR/P/2019/63 (May 9, 2019) requires all SEBI-registered Mutual Funds and Asset Management Companies to submit a quarterly AI/ML system inventory to AMFI within 15 calendar days of each quarter end. AMFI consolidates these submissions and reports to SEBI within 30 days. The inventory must cover all AI/ML applications and systems used in: portfolio management (stock selection models, factor-model automation, portfolio optimisation), risk management (VaR models, stress testing), client-facing services (fund recommendation engines, chatbots, robo-advisory features on AMC apps and MF Central), and trading algorithms. Every Indian AMC — SBI Mutual Fund, HDFC Mutual Fund, ICICI Prudential MF, Nippon India MF, Kotak MF, and 40+ others — is subject to this reporting. Key operational compliance points: the 15-day quarter-end deadline is strict; systems added or materially changed during the quarter must be included; and the circular covers AI/ML systems provided by third-party vendors as well as internally developed ones. Asset Management Companies that fail to report, report incompletely, or report inaccurately face regulatory scrutiny from SEBI. The reporting obligation has been in force since Q3 FY2020 — meaning AMCs have submitted over 20 consecutive quarterly reports by 2026 — yet no SEBI guidance has defined what level of detail constitutes adequate system description, what counts as a 'material change,' or what happens when a vendor's AI system cannot be adequately described by the AMC because the vendor treats the model as proprietary.

What is the explainability void in SEBI's mutual fund AI circular and why does it matter for investor protection?

The SEBI Mutual Fund AI/ML Circular explicitly acknowledges that 'most AI/ML systems are black boxes and their behaviour cannot be easily quantified,' and warns that investor-facing AI must not constitute misrepresentation under SEBI regulations. However, the circular creates this governance paradox: it recognises AI's explainability problem while prescribing no remedy. The result is a seven-year inventory of AI systems with no corresponding explainability architecture. The investor protection concern is specific and concrete: when an AMC's app or the MF Central platform presents an AI-generated fund recommendation to a retail investor — showing a specific fund as 'recommended for you' based on a model the AMC cannot fully explain — four SEBI obligations simultaneously apply: (1) SEBI (Mutual Funds) Regulations 1996 Regulation 18 requires that trustees ensure management of fund assets is in the interest of unitholders; an AI that systematically recommends higher-expense-ratio funds (which are more profitable for the AMC) due to training incentives violates this; (2) SEBI Circular on Risk-o-Meter (SEBI/HO/IMD/DF3/CIR/P/2020/197) requires risk level disclosure calibrated to actual portfolio risk; AI-driven portfolio construction that shifts risk exposure beyond the stated risk-o-meter level without disclosure creates a consumer harm; (3) SEBI's Advertising Code for Mutual Funds prohibits misleading claims; an AI recommendation that a retail investor reasonably interprets as personalised financial advice — but is actually a generically optimised model output — may constitute misleading communication; (4) SEBI's June 2025 consultation paper on AI/ML in securities markets explicitly proposes mandatory explainability, audit trail, and model governance requirements. AMCs that have not built explainability infrastructure while operating investor-facing AI will face a compliance gap when the consultation paper becomes a circular.

What third-party AI vendor governance obligations exist for Indian AMCs under SEBI regulations?

Indian AMCs use AI from multiple third-party vendors: Bloomberg and FactSet AI analytics, external quantitative strategy providers, fintech platforms licensed to provide robo-advisory infrastructure, and cloud AI services from AWS, Google, and Microsoft. The SEBI 2019 circular requires AMCs to include third-party AI in their quarterly inventory — but creates no further governance obligation for the AMC with respect to vendor AI. Three specific governance gaps: (1) Vendor model documentation: the AMC must report that a third-party AI system exists, but has no obligation to obtain or review the vendor's model documentation, validation history, or performance reports. An AMC relying on an opaque vendor AI for portfolio construction satisfies SEBI reporting while carrying undocumented model risk; (2) Vendor performance monitoring: there is no SEBI requirement for AMCs to monitor the ongoing performance of vendor AI systems — an AMC can deploy a vendor's fund selection AI, include it in the quarterly report, and have no contractual right to receive performance metrics or model change notifications from the vendor; (3) Vendor change notification: when a vendor materially updates their AI model — retraining on new data, changing architecture, adjusting optimization target — the AMC's deployed AI has changed in ways that may affect fund outcomes, yet the AMC may have no contractual mechanism to detect this change, let alone assess whether it constitutes a material change requiring SEBI report update. The RBI's MRMF 2026, by contrast, explicitly requires model owners to govern third-party model changes. SEBI's mutual fund AI regime has no equivalent.

How does SEBI's June 2025 consultation paper change the AI governance outlook for Indian mutual funds?

SEBI's June 2025 consultation paper on AI/ML in securities markets signals that the 2019 circular's inventory-only approach will be replaced with a comprehensive governance framework. Based on the consultation paper's proposals, AMCs face four specific upcoming obligations: (1) Mandatory explainability: AI models used in investor-facing contexts (fund recommendations, portfolio rebalancing communications, risk assessments presented to clients) must be capable of producing human-understandable explanations for their outputs — a requirement that will require AMCs to either add explainability layers to existing black-box models or replace them; (2) Pre-deployment validation: AI models used in portfolio management and trading must undergo internal validation before production deployment — AMCs that have been deploying third-party AI without internal validation will need to establish a validation function or engage external validators; (3) Audit trail: AI decisions affecting fund portfolio composition or investor-facing communications must be logged with sufficient detail to enable post-hoc review — AMCs using AI in portfolio construction must implement model logging infrastructure; (4) Human oversight: the consultation paper proposes that AI models above certain impact thresholds must have defined human oversight mechanisms — fund managers must be able to review and override AI portfolio recommendations with documented rationale. AMCs that begin building governance infrastructure in 2026 — before the consultation paper becomes a binding circular — will have a significant compliance advantage over those waiting for the final rule.

AI Governance · SEBI · Mutual Fund · AMC · AI/ML Reporting Circular · India BFSI

SEBI Mutual Fund AI ML Reporting Circular 2019 AI Governance Gap: AMC Quarterly Inventory Obligation, Portfolio AI Explainability Void, Third-Party Vendor Gap and SEBI 2025 Consultation Paper India

SEBI's 2019 AI/ML Reporting Circular (SEBI/HO/IMD/DF5/CIR/P/2019/63) created a quarterly inventory obligation for all 44 Indian AMCs — but prescribed no explainability standard, no model validation requirement, no third-party AI vendor accountability, and no audit trail. Seven years of quarterly reports have built a data inventory with no governance architecture. SEBI's June 2025 consultation paper now signals prescriptive AI governance for mutual funds is imminent. Every Indian AMC — SBI MF, HDFC MF, ICICI Prudential MF, Nippon MF, Kotak MF — faces a governance build-out before the circular lands. Zero global competitor coverage.

Zytra Research Team · June 2025 · 11 min read

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

SEBI Circular SEBI/HO/IMD/DF5/CIR/P/2019/63 (May 9, 2019) was India's first dedicated AI/ML regulatory circular for any financial sector. It mandated that every registered mutual fund submit quarterly AI/ML system inventories — an obligation that has generated 20+ consecutive quarterly reports from every Indian AMC since FY2020. Yet the circular created a governance paradox: it acknowledged AI's explainability problem in its own text while prescribing no mechanism to address it.

The result in 2026: India's mutual fund industry operates the deepest AI deployment in Indian capital markets — portfolio construction AI, trading algorithms, investor-facing recommendation engines on MF Central and AMC apps, risk management models — under a regulatory framework that inventories AI but governs none of it. SEBI's June 2025 consultation paper signals this is about to change.

The Quarterly Reporting Obligation: What It Requires

The circular's reporting cadence is strict: within 15 calendar days of each quarter end, every AMC submits its AI/ML system inventory to AMFI; AMFI consolidates and reports to SEBI within 30 days. The inventory covers all AI/ML applications — internally developed and third-party — used in portfolio management, risk management, client-facing services, and trading. No AMC is exempt: SBI Mutual Fund, HDFC Mutual Fund, ICICI Prudential MF, Nippon India MF, Kotak MF, and 40+ others all file quarterly.

What the circular does not define: what level of technical detail constitutes adequate system description, what constitutes a "material change" requiring inclusion of an updated system in the current quarter's report, or what governance the AMC must have over a third-party vendor's AI that it reports but cannot interrogate. These omissions have created 20+ quarters of filing practice built on assumptions that vary across AMCs — making the current reporting regime as much a compliance risk as a compliance mechanism.

The Explainability Void

The circular's own language frames the problem: it explicitly states that "most AI/ML systems are black boxes and their behaviour cannot be easily quantified." Having named the problem, the circular prescribed no solution. Seven years later, Indian AMCs operate investor-facing AI — fund selection models, personalised communication engines, portfolio risk assessors — under a framework that acknowledged explainability as a requirement but never mandated it.

The specific investor protection risk is multi-layered. SEBI's Advertising Code for Mutual Funds prohibits misleading communications: an AI-generated fund recommendation that a retail investor reasonably interprets as personalised advice, but is actually an optimized model output that cannot be individually explained, sits in a legal grey zone. SEBI (Mutual Funds) Regulations 1996 Regulation 18 requires that trustees ensure management is in unitholders' interests — a requirement that becomes untenable when the portfolio construction AI cannot be reviewed by trustees for alignment with unitholder interests.

Third-Party AI Vendor Gap

AMCs using vendor AI — Bloomberg Analytics, FactSet AI, external quant strategy providers, cloud AI services — must include these systems in their quarterly inventory. But the circular creates no obligation for the AMC to obtain vendor model documentation, monitor ongoing vendor model performance, or receive notification of vendor model changes. An AMC can satisfy SEBI's reporting requirement while carrying entirely undocumented model risk from a vendor whose model may have been retrained, restructured, or repurposed without the AMC's knowledge.

SEBI 2025 Consultation Paper: What's Coming

SEBI's June 2025 consultation paper on AI/ML in securities markets proposes four specific obligations that will create the governance architecture the 2019 circular omitted: mandatory explainability for investor-facing AI, pre-deployment validation for portfolio and trading AI, audit trail requirements for AI decisions affecting portfolio composition or investor communications, and human oversight mechanisms with documented override capability for high-impact AI models.

AMCs that build governance infrastructure in 2026 — explainability frameworks for investor-facing AI, internal validation capability for portfolio AI, audit logging, fund manager override documentation — will absorb the new circular with minimal disruption. AMCs that wait face a simultaneous build-out across all four dimensions when the circular is finalised.

AMC AI Governance Checklist

☐ Quarterly inventory accuracy: AI/ML system registry with material change criteria defined; 15-day quarter-end filing calendar; third-party AI systems included with vendor documentation status noted
☐ Investor-facing AI explainability: explainability layer on fund recommendation AI; ability to generate investor-readable explanation for any recommendation; SEBI Advertising Code alignment review for AI-generated communications
☐ Portfolio AI model validation: internal validation function or external validator engaged for portfolio construction and trading AI; pre-deployment validation protocol documented; ongoing performance monitoring cadence
☐ Third-party vendor AI governance: vendor AI contract includes model change notification, performance reporting rights, model documentation access; AMC-side monitoring of vendor AI outputs; material change assessment protocol
☐ Audit trail: AI decisions in portfolio construction and investor-facing communications logged; log retention period aligned with SEBI inspection requirements; log access for compliance and internal audit
☐ Human oversight: fund manager review and override capability for AI portfolio recommendations; override documentation protocol; SEBI 2025 consultation paper readiness gap assessment completed
☐ SEBI examination readiness: quarterly report accuracy documentation, model validation status, audit log, vendor governance evidence — packaged for SEBI inspection