The AI in Business Podcast Governing AI for Fraud, Compliance, and Automation at Scale - with Naveen Kumar of TD Bank
8 snips
Dec 17, 2025 Naveen Kumar, Head of Insider Risk, Analytics, and Detection at TD Bank, dives into the complexities of governing AI in the banking sector. He discusses overcoming foundational challenges like data leakage and hallucinations while advocating for role-based AI access. Naveen shares essential principles for strong AI governance, emphasizing full data visibility and treating AI agents like employees. He balances innovation with regulatory obligations and suggests practical steps for building a secure AI foundation.
AI Snips
Chapters
Transcript
Episode notes
Key Risks Slowing AI Adoption
- Data leakage, prompt injection, model inversion, shadow AI, hallucinations and model drift are major adoption blockers in banking.
- These risks require treating AI governance as a core operational concern, not an optional add-on.
Block Prompt Injection With Immutable Rules
- Defend against prompt injection by enforcing immutible rules that models cannot break.
- Prevent social-engineering-style prompts from exposing sensitive internal details.
Hallucinations Arise From Overpressure
- Hallucinations often stem from model pressure to produce answers without admitting uncertainty.
- Limiting models to role-specific context reduces irrelevant or invented outputs.