Solytics Partners Concludes the AI & Model Governance Conclave 2025, Advancing India’s Responsible AI and MRM Landscape

Scaling Responsible AI Governance & Model Risk Globally: Solytics Partners AI & MRM Conclave - Mumbai 2025

On November 25, 2025, Solytics Partners hosted the AI & Model Governance Conclave at Sofitel, BKC, Mumbai. The event brought together senior leaders from banks, NBFCs, insurers, GCCs, and industry bodies to tackle a core question:

How can Indian financial institutions scale AI and GenAI responsibly, while preserving control over risk, fairness, and accountability?

Set against the backdrop of RBI's FREE-AI principles, SEBI's AI/ML guidelines, and global frameworks such as the EU AI Act, NIST AI RMF, SR 11-7, and PRA/ECB guidance, the Conclave focused on turning regulatory expectations into practical AI governance frameworks, model risk processes, and tooling for AI risk management and model risk governance.

The half-day program combined interactive workshops, panel discussions, and networking—centered on how Model Risk Management (MRM), AI governance, and machine learning governance must evolve for the next wave of AI adoption in Indian financial services. 

1. Workshops on Governance and Validation

Workshop 1 – Governance of AI and Model Risk in Regulated Environments

Speakers: Prithivi P (Head of Technology & Innovations) and Sapna Johari (Senior Product Manager)

This session demonstrated how to move from traditional, model-centric MRM to enterprise AI governance and AI model governance that covers the full machine learning lifecycle and ML model lifecycle.

Key messages:

From model registry to AI asset registry 
Institutions need to govern complete AI assets, use cases that combine models (ML and LLMs), data pipelines, ground truth, guardrails, and infrastructure - rather than just individual models. This has direct implications for model inventory management, model lifecycle management, ai model oversight, and MLOps governance.

Four pillars of enterprise AI governance 
A simple, actionable AI governance framework (and model governance framework) was presented:

• Data Governance – data quality, lineage, ground truth, consent, and production telemetry, enabling strong data governance for machine learning (data governance machine learning).
• Model / AI Asset Governance – AI asset inventory, risk tiering, validation, and monitoring, forming the core of ML model governance, model risk governance, and model approval workflows.
• Infrastructure – secure runtime, policy-as-code (e.g., OPA), guardrails for LLMs and agents, supporting scalable responsible AI governance.
• Governance & Audit – AI Governance Council, embedded controls, evidence vaults, and dashboards, enabling explainable AI governance and regulator-ready model risk management frameworks.

AI asset lifecycle and lifecycle gates 
A structured lifecycle was discussed: discovery and classification, regulatory tagging, risk tiering, control mapping, validation, deployment, monitoring, and change management. This approach supports RBI FREE-AI, aligns with global best practice, and is consistent with machine learning model lifecycle management and model development lifecycle expectations for regulated institutions.

Using MRM Vault and the broader Solytics governance ecosystem, participants saw how to orchestrate this through centralized inventories, configurable templates, and audit-ready workflows, essential capabilities of a modern model governance platform, model monitoring platform, and model risk management software.

2. Workshop 2 – Validation and Testing of AI/ML and GenAI Models in Practice

Speakers: Agus Sudjianto (Senior Advisor) and Kannan Venkataramanan (GenAI Validation SME)

The second workshop focused on robust, regulator-ready validation for modern AI and GenAI systems, particularly LLM-based use cases, an increasingly critical part of model validation management and ML model management platforms.

Highlights:

Interpreting and stress-testing modern models 
Validation must go beyond point performance to identify weaknesses and failure modes under drift, edge cases, adversarial prompts, and domain shifts, strengthening both model lifecycle management and AI risk management processes.

Bias, fairness, and information leak 
The session explored practical fairness testing across cohorts and detecting “information leaks,” where demographic or sensitive attributes influence outcomes via proxies, and how feature engineering and debiasing mitigate this. These practices are foundational to responsible AI governance and explainable AI governance in high-stakes use cases.

Guardrails and multi-layer filtering for LLMs 
Every LLM or agent should be protected by layered guardrails—prompt sanitization, toxicity and PII filters, and policy checks—treated as formal governance controls rather than UI add-ons. This is increasingly expected in modern AI compliance platforms and AI model governance frameworks.

Automation plus human oversight 
For high-risk applications, neither human review alone nor LLM-based semantic checks are sufficient. Institutions need structured test suites, automation, and targeted human oversight working together creating a scalable pattern for MLOps governance and continuous model monitoring platforms.

These concepts were demonstrated using Nimbus Uno and MoDeVa, Solytics' platforms for ML and GenAI validation, covering LLM fine-tuning, hallucination and jailbreak testing, grounding and RAG evaluation, fairness analysis, and explainability. AI Mate was also showcased as an assistant that can automatically generate a large portion of model and AI documentation directly from Nimbus outputs, enabling model documentation automation and easing documentation requirements across the machine learning model lifecycle.

3. Panel Discussions: Priorities and Ground Realities

Panel 1 – The Future of AI Risk & Model Governance in India

The first panel, “The Future of AI Risk & Model Governance in India – 2026 Priorities,” brought together leaders from major Indian banks, global custodians, and GCCs alongside Solytics advisors.

Key insights:

Convergence of Indian and global standards 
Panelists highlighted how RBI FREE-AI and SEBI guidance are increasingly aligned with frameworks such as the EU AI Act, NIST AI RMF, SR 11-7, and PRA/ECB expectations—pushing firms to modernise their model risk management frameworks, AI governance frameworks, and model governance platforms.

From policy documents to operating models 
Institutions are now embedding governance into:

• Central AI/model inventories and consistent risk tiering, supported by strong model inventory management.
• Standardized validation toolkits for ML and LLMs, underpinning ML model governance and model validation management.
• Continuous monitoring and policy-as-code enforcement, enabled by integrated model monitoring platforms and AI governance capabilities.

India as a hub for AI governance and validation 
GCC leaders emphasized India's growing role as a global center of excellence for AI validation and model risk, supporting portfolios across multiple regions and strengthening global AI model oversight.

Top 12–18 month priorities included building a single AI/model inventory, standardizing validation and monitoring for ML and GenAI, forming AI Governance Councils that unite Risk, Data, Technology, Compliance, and Audit, and moving towards telemetry-driven continuous monitoring as part of end-to-end model lifecycle management.

Panel 2 – AI Risk & Model Governance Beyond Banking

The second panel, “AI Risk & Model Governance Beyond Banking – Building Trust, Transparency, and Accountability Across Sectors,” extended the conversation to NBFCs and cross-industry use cases, with perspectives from non-bank lenders and NASSCOM.

Key themes:

Balancing innovation and control in NBFCs 
NBFCs are deploying AI across lending, onboarding, and collections, and now must embed FREE-AI aligned model risk governance and responsible AI governance without slowing growth.

Standardization and capability building 
NASSCOM's perspective underscored the need for shared frameworks and playbooks, plus systematic capability building across risk, technology, and product teams, so institutions don't each reinvent their own model governance frameworks, machine learning governance practices, or AI governance frameworks from scratch.

Scaling governance through automation 
Panelists stressed replacing scattered spreadsheets with central AI registries, shifting from manual checks to automated validation and dashboards, and using AI for early warning on portfolio and operational risk—patterns that align with modern model governance platforms, ML model management platforms, and model risk management software.

Across both panels, there was clear consensus: the next phase of responsible AI adoption in India will be led by institutions that can blend innovation with structured AI governance, supported by the right tools, model risk management frameworks, and operating models.

4. Collaboration, Networking, and the Road Ahead

The Conclave closed with a networking cocktail and dinner, where attendees continued discussions on:

• Translating RBI FREE-AI, SEBI guidance, the EU AI Act, and NIST AI RMF into institution-specific AI governance roadmaps and model governance frameworks.
• Positioning India as a global hub for AI governance, model risk governance, and validation, leveraging GCC and analytics talent.
• Collaborating on common playbooks, validation standards, and shared tooling for AI assurance and AI compliance platforms.

A recurring message across the day was:

Governance must evolve in step with innovation, and automation is the bridge that will define the future of AI assurance.

Looking Ahead

As AI and Generative AI reshape financial services, organizations will need resilient, transparent, and regulator-ready AI governance and model risk management frameworks that scale with their AI ambitions and support the full machine learning model lifecycle management.

At Solytics Partners, we help institutions operationalize AI compliance and model risk governance through our unified platforms, built as an integrated AI compliance platform and model governance platform:

• MRM Vault – centralized model and AI governance, workflows, evidence, model approval workflows, and model inventory management.
• AI World – AI asset registry and lifecycle orchestration, enabling enterprise-wide AI model governance, ML model lifecycle oversight, and model lifecycle management.
• NIMBUS Uno & MoDeVa – ML and GenAI validation, robustness, fairness, explainability, and continuous monitoring, acting as a combined ML model management platform, model monitoring platform, and model validation management engine.
• AI Mate – automated, audit-ready documentation generation, streaMLining model documentation automation across the model development lifecycle.

To learn more about our Model Risk and AI Governance solutions, or to explore a tailored AI Governance Readiness workshop for your organization, visit Solytics Partners and discover how we support responsible AI governance end-to-end.