In this episode of Modern Cyber, Jeremy is joined by cybersecurity veteran Caleb Sima for a deep dive into the practical realities of securing AI inside organizations. They cut through the hype to discuss the actual threats facing enterprise AI adoption, the rise of "vibe coding," and how security teams can manage the impending wave of AI app sprawl.

Key Episode Highlights:

• The Core Threats: Caleb identifies prompt injection as the number one most likely and impactful threat model for AI systems today, followed closely by data poisoning.
• The Rise of "App Sprawl": As employees across departments like HR and Finance use AI to build their own functional applications, organizations will face a massive shadow IT challenge without proper deployment pipelines.
• Defending the Inputs and Outputs: Managing AI security requires an approach similar to handling cross-site scripting, monitoring the inputs coming from untrusted sources and analyzing the outputs to prevent unauthorized actions.
• Getting Back to Basics: To secure AI, organizations must start with foundational visibility, establishing AI councils, and routing all LLM traffic through centralized enterprise gateways or firewalls.

About Caleb

Caleb is a multi-time founder, CEO and CTO, and also a CISO and practitioner at CapitalOne, DataBricks and RobinHood. Caleb has also recently started his own cyber investment firm, WhiteRabbit. At his core, Caleb is an engineer who loves problem-solving, getting into the weeds at the keyboard, and building things that matter.

Episode Links

• Caleb Sima on LinkedIn: https://www.linkedin.com/in/calebsima/
• WhiteRabbit: https://wr.vc/