India's insurance sector has been deploying AI for years. Claims triage, underwriting risk scoring, fraud detection, customer service chatbots — the AI is there. What hasn't been there is a regulatory framework defining how it should be governed. IRDAI's June 2026 working group changes that.
The working group's significance extends beyond insurance: its composition — with ReBIT (RBI's technology arm) and CERT-In alongside IIIT Hyderabad and insurance company representatives — signals that India's sectoral AI governance frameworks are being coordinated at the regulatory level. The same institutions that built and advised RBI's MRMF and the CERT-In Blueprint are now shaping the insurance framework. That convergence is deliberate. India's AI governance architecture is being built as one consistent system across sectors, not as four independent regulatory silos (RBI, SEBI, IRDAI, MeitY) arriving at incompatible standards.
Working Group's Eight-Point Mandate: What It Will Produce
AI Adoption Assessment
How far have Indian insurers already deployed AI? The working group will map current AI deployment: which products, which use cases, which vendors, which models. This is the AI inventory for the insurance sector at the regulatory level. It will likely produce the same picture RBI FREE-AI produced for banking: concentrated deployment in fraud and underwriting, limited governance infrastructure around it.
Governance Maturity Evaluation
Of the AI that is deployed, how mature is the governance around it? IRDAI will assess whether insurers have model inventories, validation processes, explainability documentation, board oversight, and risk tiering. The maturity assessment will inform how prescriptive the final guidelines need to be — if governance maturity is low, more detailed requirements will be issued.
Global Regulatory Review
IRDAI will review EU AI Act (insurance as high-risk application), UK FCA's insurance AI approach, MAS FEAT framework for Singapore's insurers, IOSCO's approach to AI in financial services, and NAIC's AI principles for US insurance. The global scan will inform IRDAI's choices between principles-based and prescriptive approaches, risk tiering methodology, and explainability requirements.
Ethical and Explainable AI Frameworks
Underwriting AI and claims adjudication AI both affect policyholders directly. The Understandable by Design sutra (MeitY/FREE-AI) requires that AI decisions in these contexts be explainable to affected policyholders. An insurer that denies a claim via AI or prices a life insurance policy via AI must be able to explain the decision — in a language and at a level of detail the policyholder can understand, not just in model documentation language a data scientist can audit.
AI Cybersecurity Risk Assessment
CERT-In's presence on the working group ensures this mandate gets treated seriously. AI systems in insurance are attack targets: claims processing AI can be manipulated to approve fraudulent claims; underwriting AI can be attacked to produce favorable risk scores; customer service AI can be jailbroken to provide misleading policy information. CERT-In Blueprint's principles (AIBOM, adversarial testing, kill switches) will likely translate into insurance-sector cybersecurity AI requirements through this mandate.
AI Risk Stress Testing
Insurance AI stress testing has two dimensions unique to the sector: (1) model failure at the portfolio level — if an underwriting AI misprices a systemic risk (e.g., climate risk) across thousands of policies simultaneously, the resulting claims concentration could threaten insurer solvency. This is analogous to model risk in banking but with potentially faster and larger impact per event; (2) correlated AI failure across insurers — if multiple Indian insurers use the same vendor AI (common in the India market where CBS vendor concentration is high), simultaneous model failure could create systemic risk analogous to FSB's concentration risk concern for banking AI.
AI Audit Framework
IRDAI will develop an AI audit framework for insurance — what an IRDAI examiner will review when conducting an AI governance examination of an insurer. This directly mirrors what RBI MRMF created for banking (model inventory, validation evidence, board governance, audit trail). Insurers that build AI governance programs now against the RBI MRMF/FREE-AI template — adapted for insurance use cases — will require minimal additional work when the IRDAI audit framework is finalized.
Capacity Building Requirements
India's insurance sector has AI skill distribution similar to its banking sector: large PSU insurers and major private insurers have data science teams; smaller insurers and regional players depend almost entirely on vendor AI and CBS-embedded AI. The capacity building mandate will likely produce differentiated requirements by insurer size, analogous to RBI FREE-AI's Pillar 2 (Capacity Building) and UCB tier differentiation — proportional governance for proportional AI deployment sophistication.
Five Insurance AI Use Cases in Governance Scope
Underwriting AI
Risk assessment and premium pricing for life, health, general. Governance requirements: India-specific fairness testing (disability, religion, caste proxy); DPDP Act consent for health data used in underwriting AI; explainability of premium pricing decisions; accuracy validation against claim outcomes; human override capability for AI-flagged rejections.
Claims Adjudication AI
Automated claims assessment, medical claim validation, motor claims AI. Governance: adverse decision explanation mandatory for AI-driven claim denial; false positive audit (legitimate claims incorrectly flagged as fraudulent); policyholder appeal process for AI decisions; kill switch for claims AI that produces systematic errors. DPDP Act applies to medical data in health claims AI.
Fraud Detection AI
Insurance fraud is significant (staged accidents, inflated medical claims, ghost policies). Governance: bias testing for false positive rates across geographic/community segments; model drift monitoring as fraud patterns evolve; explanation capability for claim investigations; adversarial testing against known fraud-injection techniques. CERT-In Blueprint adversarial testing mandate applies to fraud AI as an attack surface.
Customer Service AI
Chatbots for policy queries, AI-assisted sales, complaint resolution. INDIC-BIAS testing required for customer service LLMs (85 identity groups, language/script proxy discrimination). Policy explanation AI must be accurate — incorrect information about policy terms, exclusions, or claim procedures constitutes mis-selling. Multilingual Indian language support quality must be equitable across languages, not English-first with degraded regional language service.
Investment and Actuarial AI
Investment portfolio AI, actuarial mortality/morbidity modeling, catastrophe AI, climate risk pricing AI. Governance: solvency impact stress testing; IRDAI investment guidelines compliance for AI-assisted investment decisions; third-party model validation for actuarial AI; climate risk AI specifically under RBI climate disclosure alignment (CRIS integration for insurers as well as banks). Model failure at portfolio level = solvency risk.
Build Now Against Seven Sutras + MRMF Template — Adapt Later When IRDAI Issues Final Framework
The IRDAI working group's three-month timeline means guidelines could issue Q4 2026 or Q1 2027. Insurers that wait will face a compressed implementation timeline. The convergent architecture of India's AI governance frameworks means building against the existing frameworks now is a sound investment:
- AI system inventory: Build a model inventory across all five use case categories now. IRDAI's governance maturity assessment will ask for this. Insurers that can present a complete inventory — model names, vendors, use cases, data inputs, decision outputs, customer impact — will receive substantially different examiner attention than those starting from scratch.
- Policyholder explainability: Underwriting AI and claims AI must produce customer-intelligible explanations for adverse decisions before the framework issues. IRDAI's Working Group has "ethical and explainable AI" as an explicit mandate — the final framework will require explainability. Build now: adverse action explanations for underwriting rejections and claim denials, in the policyholder's preferred language.
- DPDP Act data governance for AI: Health insurance underwriting AI uses sensitive personal data (health records, family medical history). DPDP Act 2023 requires consent for this processing and applies non-discrimination protections. Map your underwriting AI inputs against DPDP Act sensitive data categories now — before IRDAI adds insurance-specific data governance requirements on top.
- CERT-In Blueprint alignment for insurance AI: CERT-In is on the IRDAI working group. CERT-In Blueprint's AI cybersecurity requirements (AIBOM, adversarial testing, kill switches) will translate into insurance AI requirements. Build your insurance AI AIBOM now: document every AI system, its vendor, its data inputs, its attack surface. This satisfies CERT-In requirements whether IRDAI specifically mandates it or not.
- Climate risk AI for insurers: India's climate risk AI governance is entering insurance through two channels: IRDAI's own mandate (catastrophe modeling, climate risk underwriting) and RBI's climate disclosure framework (which insurers holding bank accounts and investments must also navigate). Climate risk AI governance built for RBI compliance will also satisfy IRDAI's climate AI requirements when they issue.