Protegrity on LLMs in Banking: Securing Sensitive Data for AI Adoption

By Protegrity

Dec 19, 2025

Summary

5 min

LLMs in banking need data-centric security to scale:
In an E-ChannelNews interview, Protegrity CTO Sameer Tiwari explains how banks can adopt LLMs without exposing PII, PCI, or PHI—by protecting sensitive data at the core so AI workflows stay secure even if perimeter defenses fail.
Secure AI workflows with governance, guardrails, and privacy tooling:
Protegrity AI Team Edition supports responsible GenAI in regulated environments using discovery and classification, anonymization, synthetic data, and semantic guardrails—enabling auditable, compliant AI adoption across prompts, training, and outputs.

In a recent E-ChannelNews interview, Sameer Tiwari, Chief Technology Officer at Protegrity, shared how the company helps banks and healthcare providers protect sensitive data while advancing AI initiatives. He emphasized a data-centric approach—protecting information at its core so it remains secure even if other defenses are breached.

Key themes from the interview

LLMs in regulated environments: Banks see the upside of large language models, but concerns about data exposure often drive strict security controls that can limit effectiveness.
AI Team Edition capabilities: Tiwari outlined how Protegrity AI Team Edition supports secure AI workflows with anonymization, discovery, synthetic data, and semantic guardrails—helping teams manage data responsibly while addressing legal and compliance needs.
Governance for agentic systems: He underscored governance as a foundational requirement for data management—especially as agentic systems introduce new risks, including agents creating other agents.
Quantum-ready security posture: Tiwari discussed the implications of quantum computing for cybersecurity, including the need to rotate encryption keys regularly and adopt quantum-safe algorithms.

Supporting developers and accelerating adoption

The discussion also touched on Protegrity’s approach to enabling teams with accessible tools and feedback loops, including a free development edition shared publicly to encourage collaboration and user input—while keeping pace with evolving cyber threats.\

Read the full article