What are the five CSCRF categories and what AI cybersecurity obligations apply to each?

SEBI CSCRF (SEBI/HO/ITD-1/ITD_CSC_EXT/P/CIR/2024/113, August 20, 2024) classifies all SEBI regulated entities into five categories based on scale, trading volume, AUM, and client count, with escalating AI cybersecurity obligations: Category 1 (Market Infrastructure Institutions): NSE, BSE, MSE, all clearing corporations, NSDL, CDSL — the highest tier. Category 1 entities face the most comprehensive AI cybersecurity requirements: mandatory Security Operations Centre (SOC) with AI/ML-powered threat detection and anomaly monitoring; AI-based vulnerability assessment as expected standard of care (formalized in the May 2026 advisory); dedicated AI cybersecurity incident response procedures; comprehensive AI vendor security governance; board-level AI cybersecurity reporting. Category 2 (Qualified Regulated Entities): large brokers, AMCs above defined AUM thresholds, large depositories participants, qualified foreign investors. Category 2 entities must maintain SOC capabilities with AI-assisted threat detection; implement AI-based vulnerability scanning for critical applications; maintain authorized AI cybersecurity tool lists; conduct annual AI cybersecurity audits. Category 3 (Mid-tier entities): medium-sized brokers, mid-tier AMCs, investment advisers above defined client thresholds, portfolio managers above defined AUM thresholds. Category 3 entities must implement AI-assisted cybersecurity monitoring (SOC or equivalent); conduct periodic AI cybersecurity risk assessments; maintain AI tool governance for cybersecurity applications. Category 4 (Smaller regulated entities): small stockbrokers, smaller IAs and RAs, small portfolio managers, smaller AIFs. Category 4 entities must implement baseline cybersecurity controls; AI governance obligations are proportionate — basic authorized AI tool governance and cybersecurity policy coverage of AI systems. Category 5 (Smallest entities): micro-sized regulated entities below all category thresholds. Category 5 entities must implement minimum cybersecurity hygiene; AI governance is addressed through basic policy rather than technical controls. Compliance timeline: Category 1 entities — full CSCRF compliance required from the circular's effective date. Categories 2–5 had phased compliance timelines extending through 2025.

What is the Principle of Equivalence in SEBI CSCRF 2024 and why does it matter for banks regulated by both RBI and SEBI?

The December 31, 2024 clarification circular (SEBI/HO/ITD-1/ITD_CSC_EXT/P/CIR/2024/184) introduced the Principle of Equivalence, which is particularly significant for dual-regulated entities — institutions supervised by both RBI and SEBI. The Principle of Equivalence states that CSCRF compliance applies only to SEBI-regulated systems and operations (the Principle of Exclusivity), and that for dual-regulated entities, compliance with the applicable cybersecurity framework of the other regulator (RBI's IT Governance Master Directions) satisfies the equivalent CSCRF requirement for the overlapping system — provided the other regulator's requirement is equivalent in substance. Four practical implications for dual-regulated entities: (1) Commercial banks that are also SEBI registered intermediaries (for example, a bank that operates a depository participant business, or a bank that is a SEBI-registered stockbroker through its securities arm) are dual-regulated. The Principle of Equivalence means they can demonstrate CSCRF compliance for their banking systems through RBI cybersecurity compliance, rather than running two parallel cybersecurity compliance programs for the same infrastructure; (2) The Principle does not exempt dual-regulated entities from CSCRF — it aligns the compliance standard. Where a bank's SEBI-regulated function uses infrastructure that is also RBI-regulated, the bank must demonstrate that its RBI cybersecurity compliance covers the relevant SEBI requirements. If the RBI framework is silent on a CSCRF requirement (e.g., a specific AI SOC requirement that RBI has not yet mandated), the CSCRF requirement applies independently; (3) AI cybersecurity governance at dual-regulated entities must be mapped separately for the SEBI-regulated system component and the RBI-regulated system component. The AI SOC requirement under CSCRF Category 2 may or may not have a direct RBI equivalent — compliance teams must do the mapping rather than assuming blanket equivalence; (4) The December 2024 clarification also clarified that CSCRF supersedes and replaces all prior SEBI cybersecurity circulars — entities that had built compliance programs around earlier SEBI cybersecurity guidance (the 2018 framework, specific product-line cybersecurity requirements) must migrate to the CSCRF architecture.

How does SEBI CSCRF 2024 relate to the SEBI AI Vulnerability Detection Advisory 2026 and the SEBI AI/ML Responsible Usage Framework?

Understanding the relationship between these three SEBI instruments is essential for compliance teams building an integrated SEBI AI governance program: (1) SEBI CSCRF 2024 (parent framework): the CSCRF is the foundational cybersecurity governance framework for all SEBI regulated entities. It establishes the five-tier classification, the SOC mandate, the vulnerability assessment requirements, and the AI cybersecurity tool governance obligations. The CSCRF creates the compliance architecture within which subsequent AI-specific instruments operate. Compliance owner: CISO and technology risk team; (2) SEBI AI Vulnerability Detection Advisory 2026 (child instrument under CSCRF): the May 2026 advisory (SEBI HO/13/19/12(1)2026-ITD-1_CIMGI/10873/2026) is a specific CSCRF-anchored advisory that operationalizes the AI-based vulnerability assessment expectation — naming AI-based VA tools as 'expected standard of care', mandating SBOM, and specifying M-SOC onboarding timelines. It is a CSCRF implementation guidance document, not a standalone obligation. It amplifies and specifies what 'AI-based VA' means under the CSCRF parent; (3) SEBI AI/ML Responsible Usage Framework (parallel track): the Responsible Usage Framework (consultation June 2025, final circular August 2026) governs AI/ML used in business operations — trading decisions, risk management, client advisory, portfolio management. It is not a cybersecurity instrument. The Framework's compliance owner is the business/compliance function, not the CISO. The Framework and CSCRF together form the two AI compliance tracks: operational AI (Responsible Usage Framework) and cybersecurity AI (CSCRF + 2026 advisory); (4) Two-track compliance architecture: a SEBI regulated entity has two parallel AI compliance programs — operational AI compliance under the Responsible Usage Framework (what AI does in business) and cybersecurity AI compliance under CSCRF (how AI is used to protect the business). Many entities manage these as one program, which creates compliance gaps when the ownership, documentation, and evidence requirements are different between the two tracks.

What specific SOC and AI cybersecurity tool governance obligations does SEBI CSCRF create for Category 1 and Category 2 entities?

SEBI CSCRF's SOC and AI cybersecurity tool governance obligations for Category 1 and Category 2 entities create five specific requirements: (1) SOC mandate with AI/ML-powered capabilities: Category 1 and Category 2 entities must maintain a Security Operations Centre (or engage an equivalent M-SOC under the SEBI M-SOC framework). The SOC must incorporate AI/ML-based threat detection and anomaly monitoring capabilities — the CSCRF's continuous security monitoring requirement cannot be satisfied by legacy rule-based SIEM-only monitoring for entities at this tier. The SEBI AI Vulnerability Detection Advisory 2026 formalizes the 'expected standard of care' language that was implicit in the CSCRF's threat detection requirements; (2) Authorized AI cybersecurity tool governance: entities must maintain a formal list of authorized AI cybersecurity tools. Employees and contractors cannot use unapproved AI tools in security functions — the same governance requirement that applies to operational AI under the Responsible Usage Framework applies separately to cybersecurity AI. Unauthorized AI security tools create liability under both the Responsible Usage Framework and the CSCRF; (3) AI vendor security due diligence: AI cybersecurity tool vendors must be assessed for their own security practices — an AI security tool that has known vulnerabilities or inadequate data handling creates the same type of supply chain risk that the CSCRF is designed to address. Vendor assessment for AI cybersecurity tools is distinct from due diligence on operational AI vendors under RBI's Outsourcing MD; (4) Incident response for AI-related cybersecurity events: entities must have documented incident response procedures that specifically address AI-related cybersecurity scenarios — adversarial attacks on AI models (data poisoning, model inversion), AI-accelerated phishing or credential attacks, and unauthorized access to AI model training data or inference data. Generic IT incident response plans that do not address AI-specific scenarios are non-compliant at Category 1–2 tier; (5) Board-level AI cybersecurity reporting: Category 1 entities must provide board-level cybersecurity reporting that includes AI cybersecurity risks, AI system vulnerabilities, and AI vendor security posture. Board reporting that covers only traditional cybersecurity metrics without AI-specific dimensions is incomplete under the CSCRF's governance requirements.

AI Governance · SEBI · CSCRF · Cybersecurity · SOC · Dual-Regulated Banks · Principle of Equivalence · India BFSI

SEBI CSCRF 2024 AI Governance: Five-Tier Cybersecurity Classification, SOC AI-Powered Threat Detection Mandate for Category 1–3 Entities, Principle of Equivalence for Dual-Regulated Banks and Parent AI Cybersecurity Framework for All 10,000+ SEBI Regulated Entities India

SEBI Cybersecurity and Cyber Resilience Framework (SEBI/HO/ITD-1/ITD_CSC_EXT/P/CIR/2024/113, August 20, 2024; December 2024 clarification SEBI/HO/ITD-1/ITD_CSC_EXT/P/CIR/2024/184) — broadest-scope SEBI cybersecurity instrument, superseding all prior SEBI cybersecurity circulars. Five-tier entity classification (Category 1: MIIs; Category 2: qualified REs; Categories 3–5: scaled by volume/AUM/clients) with escalating AI cybersecurity obligations. SOC mandate for Category 1–3 with AI/ML-powered threat detection and anomaly monitoring as expected capability. AI-based vulnerability assessment as standard of care. Authorized AI cybersecurity tool governance. December 2024 clarification: Principle of Exclusivity (CSCRF applies only to SEBI-regulated systems) and Principle of Equivalence (for dual-regulated banks, RBI cybersecurity compliance satisfies equivalent CSCRF requirement — with gaps still subject to CSCRF). Parent framework for SEBI AI Vulnerability Detection Advisory 2026 (child instrument). Distinct from SEBI AI/ML Responsible Usage Framework (operational AI) — CSCRF governs cybersecurity AI; different compliance owner (CISO/CTO vs. compliance team). All 10,000+ SEBI regulated entities. Zero global competitor coverage.

Zytra Research Team · August 2024 · 11 min read

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

The SEBI AI Vulnerability Detection Advisory 2026 is often discussed as the definitive SEBI AI cybersecurity instrument. It is not — it is a child instrument operationalizing the AI cybersecurity expectations that already existed in the CSCRF parent framework. Understanding SEBI AI cybersecurity compliance requires understanding the CSCRF first: the Advisory tells you what 'AI-based VA' means; the CSCRF tells you why it is required and for which entities.

The CSCRF is also the broadest SEBI circular by entity scope — it covers all 10,000+ SEBI regulated entities from Category 1 MIIs to the smallest Category 5 stockbrokers. The five-tier classification is not just an administrative convenience; it creates genuinely different compliance obligations at each tier, with AI cybersecurity requirements calibrated to the entity's systemic significance and operational scale.

The Two-Track AI Compliance Architecture

A SEBI regulated entity running an AI compliance program has two separate tracks: the Responsible Usage Framework (operational AI — what AI does in business operations, compliance owner: compliance/risk team) and the CSCRF (cybersecurity AI — how AI is used to protect the business, compliance owner: CISO/CTO). Both tracks are mandatory. Both have SEBI reporting and audit implications. Many entities conflate them — which creates gaps in both tracks because the documentation standards, evidence requirements, and oversight functions are different.

Dual-Regulated Entity Compliance Gap Analysis

The Principle of Equivalence introduced in the December 2024 clarification reduces compliance burden for dual-regulated entities but does not eliminate it. Compliance teams at banks with SEBI-regulated subsidiaries or business lines must map: which RBI cybersecurity requirement is claimed as equivalent to which CSCRF requirement; whether the RBI requirement is actually equivalent in substance; and whether any CSCRF requirement has no RBI equivalent. Where gaps exist, the CSCRF requirement applies independently. This mapping exercise is not optional — it is the proof that the Principle of Equivalence claim is valid.

SEBI CSCRF AI Compliance Checklist

☐ Category assessment: entity's CSCRF category confirmed against current SEBI classification criteria; implications for AI cybersecurity obligations documented
☐ SOC/M-SOC: Category 1–3 SOC operational with AI/ML threat detection and anomaly monitoring capability; M-SOC onboarding status confirmed per SEBI 2026 advisory timeline
☐ Authorized AI tool governance (cybersecurity): formal authorized AI cybersecurity tool list maintained; employee/contractor unauthorized AI tool prohibition implemented
☐ AI-based VA: AI-powered vulnerability assessment implemented for critical applications per CSCRF + May 2026 advisory; SBOM in place for critical application open-source stack
☐ AI cybersecurity incident response: documented procedures for AI-specific cybersecurity scenarios (model adversarial attacks, AI-accelerated threats, AI training data breach)
☐ Dual-regulated mapping (where applicable): Principle of Equivalence mapping completed; RBI cybersecurity compliance evidence matched to CSCRF equivalents; gaps under independent CSCRF compliance identified and remediated
☐ CSCRF supersession confirmed: prior SEBI cybersecurity circular compliance programs migrated to CSCRF architecture; no legacy framework obligations carried forward
☐ Two-track audit: CSCRF AI cybersecurity audit (CISO-owned) and Responsible Usage Framework audit (compliance-owned) operating independently; no conflation of scope