SEBI's May 2026 AI/ML circular creates a quarterly reporting obligation that most mutual fund compliance teams have not yet operationalized. The mechanism is AMFI-intermediated — fund houses report to AMFI rather than directly to SEBI — but the obligation originates in SEBI regulation and SEBI retains inspection authority. Asset managers who delay treating this as a routine compliance function will find themselves scrambling at quarter-end.
This page covers: what must be reported, which AI systems are in scope, how client-facing AI is treated differently from internal AI, what the August 2026 SEBI signal means for future obligations, and the vendor AI governance trap that catches fund houses using third-party AI tools.
What the Quarterly Report Must Cover
In-Scope AI/ML Applications (No Materiality Threshold)
The circular covers ALL AI/ML applications used by a mutual fund in two categories:
1. Offered to investors — robo-advisory tools, AI-driven portfolio recommendation engines, chatbots that provide fund information, AI-generated fund fact sheet summaries, goal-based planning tools.
2. Used for compliance and management — AI-powered AML/KYC systems, NAV calculation models with ML components, AI-driven surveillance for front-running detection, automated rebalancing triggers, risk scoring models.
The "no materiality threshold" language is significant: a fund house that uses a third-party AI chatbot for investor queries — even one developed externally — cannot exclude it from the inventory on grounds of small AUM or limited usage.
Required Inventory Contents (Per AI Application)
| Data Field | What SEBI Expects | Common Gap |
|---|---|---|
| Application name and description | Plain-language description of what the AI does and where it is deployed | Describing the vendor product name rather than what it does for fund operations |
| AI/ML model type | Classification, regression, NLP, recommendation, generative — with the specific algorithm or model family | "Machine learning model" without specificity |
| Data inputs | What data feeds the model — market data, fund NAV history, investor profiles, transaction data | Omitting third-party data sources the vendor uses |
| Decision or output | What decision the AI makes or what output it produces — recommendation, risk score, flag, generated text | Missing downstream use of the output |
| Human oversight mechanism | What human review exists before the AI output affects investors or compliance decisions | Asserting "human in the loop" without describing the actual process |
| Senior management sign-off | For client-facing AI: named senior manager accountable for the application | No named owner |
| Vendor details (if third-party) | Vendor name, contractual AI governance obligations, data security clauses | "Covered by vendor contract" without specifics |
Client-Facing vs. Internal AI — The Two-Tier Obligation
Client-facing AI (heavier regime):
- Mandatory bias audit before deployment and annually thereafter
- Investor disclosure: fund offer documents and client communications must disclose that AI is used and for what purpose
- Named senior management accountability
- Pre-deployment validation including shadow testing against live market data
- Post-deployment continuous monitoring with governance-connected escalation
Internal AI (lighter regime):
- Inventory inclusion required (no exemption)
- Model documentation and validation
- Monitoring for performance degradation
- No mandatory investor disclosure, but internal risk committee oversight required
The distinction matters practically: a fund house using AI to generate draft investment commentary that analysts then review is "internal AI" (lighter regime); a fund house publishing AI-generated portfolio recommendations directly to investors on its app is "client-facing AI" (heavier regime). The test is whether the AI output reaches investors, not whether a human reviews it internally.
The Vendor AI Governance Trap
The circular makes asset managers responsible for data security and regulatory compliance even when using external AI tools — liability does not transfer to the vendor. This creates three specific obligations for fund houses using third-party AI:
1. Data security: if an external AI vendor processes fund AUM data, investor portfolio data, or transaction histories, the fund house must ensure the vendor's data security posture meets SEBI requirements. A standard IT vendor contract is insufficient — an explicit AI data governance clause is required.
2. Model testing: pre-deployment validation and shadow testing requirements apply to third-party models deployed by the fund house. "The vendor validated it" does not satisfy SEBI's requirements. The fund house must conduct or commission its own validation.
3. Incident reporting: if a third-party AI tool causes a compliance failure, the fund house is the reporting entity. Vendor-caused AI incidents must appear in SEBI compliance reporting.
What August 2026 Signals Mean for 2027
On August 19, 2026, SEBI signaled imminent comprehensive AI guidelines adding:
- Mandatory human oversight requirements for ALL automated systems (not just algo trading)
- Kill-switch obligations for AI-driven investment decision systems
- Extension of the SEBI MRMF validation requirements to AMC AI models
Fund houses that have built a rigorous quarterly reporting process — with proper inventory, validated models, and documented oversight — will have a much shorter path to compliance with these forthcoming requirements than those who have treated quarterly reporting as a checkbox exercise.
Five-Step Quarterly Reporting Readiness Program
- AI System Discovery — conduct a full inventory of all AI/ML applications across investor-facing products, compliance functions, and operational processes. Include vendor-supplied tools. Target: complete before Q3 FY2026-27 quarter end.
- Classify by Tier — sort each application into client-facing (heavier regime) or internal (lighter regime). Ambiguous cases (e.g., AI used internally to generate content that is then published) default to client-facing tier.
- Assign Senior Management Owners — for every client-facing AI application, name a senior manager with explicit accountability. Document in board-approved AI governance policy.
- Validate and Shadow Test — for client-facing AI not yet validated, initiate pre-deployment validation including shadow testing against live market data. This is not optional even for AI already in production — retroactive validation is required.
- Build the Quarterly Report Template — create a standardized template covering all seven required data fields per application. Assign a compliance owner for AMFI submission within 15 calendar days of each quarter end.
Frequently Asked Questions
Which SEBI-regulated entities must comply with the AI/ML quarterly reporting circular?
SEBI Circular HO/13/19/12(1)2026-ITD-1_CIMGI/10873/2026 (May 5, 2026) addresses all SEBI-regulated entities including stock exchanges, depositories, brokers, mutual funds, custodians, credit rating agencies, merchant bankers, and portfolio managers. Mutual funds and AMCs report their AI/ML inventory to AMFI within 15 calendar days of each quarter end. The obligation has no materiality threshold — any AI/ML application offered to investors or used for compliance and management functions is in scope regardless of AUM or usage volume.
How does the SEBI mutual fund AI reporting obligation differ from the SEBI IA/RA AI disclosure requirement?
The SEBI AI/ML inventory reporting (May 2026 circular) is a regulatory reporting obligation — fund houses submit structured inventories to AMFI each quarter covering all AI systems. The SEBI IA/RA AI disclosure requirement (Circular SEBI/HO/MIRSD/MIRSD-PoD-1/P/CIR/2025/003, January 8, 2025) is a client-level disclosure obligation — investment advisers and research analysts must disclose to individual clients which AI tools they use and how. The two obligations are complementary, not alternative.
Can a mutual fund exclude a vendor-supplied AI tool from the SEBI quarterly report?
No. The circular makes asset managers responsible for regulatory compliance even when using third-party AI tools — liability does not transfer to the vendor. A mutual fund using a vendor's robo-advisory platform, AI-driven compliance tool, or investor-facing chatbot must include that tool in its quarterly AI/ML inventory and ensure the vendor's data handling meets SEBI requirements.