Home / Episodes / S13 E2: Securing Enterprise …

Season 13 — Episode 2

S13 E2: Securing Enterprise GenAI: Data Privacy & Security Guardrails with Steven Walchek, Founder & CEO of Liminal

Tuesday, September 01, 2026 • 35:23 listen

Press Play to Listen While You Browse

Code Story | Startup Podcast for Technical Founders

0:00 35:23

Steven Walchek is originally from the Bay Area in California, but lives in Denver now. He's been there for 8 years with his family. He comes from a family of risk takers and entrepreneurs - for example, his Dad was a tech guy who did his thing and had success in Silicon Valley. Outside of tech he is a family man, married with kids. He enjoys doing anything he can with them, as they are his "why". He plays hockey, coaches hockey, and rides dirt bikes and mountain bikes.

Steven and his team built an application security platform, and ended up hitting every branch on the way down from the startup tree. When they sat down to figure out what they needed to do next, they combined their deep security tooling with a multi-modal platform, to enable complete AI governance in one central hub for regulated companies.

This is the creation story of Liminal.

S13 E1: The Token Winter Is Coming: Why Code Inflation Fails to Ship Real Business Value with Dave Garcia, Founder & Co-CEO of Pensero

Aug 25, 2026

S10 E8: From Cisco Meraki and Splunk to Fixing Privilege Sprawl: Why Treating Identity as a Static One-Time Login Invites Cloud Catastrophe with Shashwat Sehgal, Co-Founder & CEO of P0 Security

Oct 28, 2024

S9 E29: Overthrowing Shelf-Ware IGA and Engineering Engineering-Led Least-Privilege Frameworks with Umaimah Khan, Co-Founder & CEO of Opal Security

Jun 18, 2024

S13 Bonus: The Intention Layer: Why Coding Agents Need Product Judgment with Drew Dillon, Founder & CEO of Brief

Aug 27, 2026

Key Takeaways

Enterprise AI Adoption Demands Centralized Governance: Regulated industries cannot safely adopt generative AI tools or AI models without a single control hub that enforces security policies, data privacy rules, and compliance guardrails across all prompts and payloads.
Preventing Sensitive Data Spills in Multi-Modal Traffic: As enterprise AI usage shifts from text to multi-modal inputs (including image uploads, PDFs, audio, and code snippets), security platforms must perform real-time redaction and inspection before data touches third-party models.
Pivoting from Failure to Product-Market Fit: After building an application security platform and "hitting every branch on the way down," Walchek’s team leveraged their core security tooling expertise to pivot directly into solving the emerging enterprise GenAI governance gap.
Balancing Employee Productivity with CISO Control: Simply blocking external AI tools creates "Shadow AI" habits; a modern security posture gives employees frictionless access to AI models while automatically filtering out PII, IP, and confidential corporate data behind the scenes.
Horizontal AI Safety Built for Regulated Markets: Highly regulated sectors—such as finance, healthcare, and legal—require customizable guardrails that allow security administrators to tune compliance policies to match specific state, federal, or industry regulations.

Frequently Asked Questions

What is Liminal, and what problem does it solve for enterprise companies?

Liminal is a multi-modal security and AI governance platform. It allows regulated enterprises to safely adopt generative AI by establishing a centralized control layer that monitors, redacts, and secures sensitive data in real time across all AI tools and APIs.

What inspired Steven Walchek and his team to launch Liminal?

After facing major obstacles with their previous application security startup, Walchek and his team realized that enterprises were struggling to govern the flood of internal generative AI usage. They combined their existing security infrastructure with a multi-modal gateway to create Liminal.

How does Liminal protect sensitive data without breaking developer or employee workflows?

Liminal acts as a security guardrail. When users interact with AI models or tools, Liminal inspects inputs in real time—automatically redacting sensitive PII, proprietary code, or financial records—before routing clean requests to the underlying AI model.

What types of data inputs can Liminal scan and secure?

Liminal handles multi-modal data streams, meaning it can scan, inspect, and enforce security policies across structured text, unstructured documents, PDFs, image attachments, and API payloads.

What is Steven Walchek’s background prior to founding Liminal?

Originally from the Bay Area and currently living in Denver, Walchek comes from a family of tech entrepreneurs. He is an experienced founder and technical leader who previously built enterprise security and application management tools.

How does Liminal help CISOs prevent "Shadow AI" within an organization?

Instead of forcing security teams to issue blanket bans on AI tools—which often drives employees to use personal accounts on unmonitored devices—Liminal provides a secure access gateway where usage is monitored, policy-enforced, and fully visible to IT administrators.