AI Red Teaming At Scale with Bret Kinsella, VP of AI Safety and Security at TELUS Digital

Bret Kinsella, VP of AI Safety and Security at TELUS Digital, is helping enterprises move from AI guardrail guesswork to systematic, scalable testing that keeps generative AI both safe and useful. In this episode, Russ and Bret dig into how TELUS Digital built Fortify, an AI safety platform that lets teams simulate thousands of attacks in hours so they can find vulnerabilities before customers or adversaries do.​They explore why traditional red teaming cannot keep up with the pace of gen AI deployment, how probabilistic models expand the attack surface, and why prevention needs to start long before launch. Bret explains how Fortify automates attack generation across hundreds of risk scenarios, maps findings to frameworks like OWASP, NIST, and MITRE, and turns security experts and product managers alike into editors of AI safety instead of manual testers. He also shares real world results from government and enterprise deployments, how TELUS scaled from billions to trillions of AI tokens, and why shadow AI, data leakage, and reputational risk are forcing organizations to rethink AI governance and trust from the ground up.​Topics Covered:[00:01] Welcome and intro, Bret Kinsella and TELUS Digital​[00:30] TELUS Digital, FuelIX, and the scale of AI across the business​[02:00] Why generative AI changes the security and risk model​[03:30] The limits of manual red teaming for gen AI systems​[05:00] Intervention vs prevention and the problem with relying only on guardrails​[07:00] Automating AI safety testing so non technical teams can run it​[09:30] Running thousands of attacks in hours and what that reveals in production bots​[11:30] Novel attacks, repeat testing, and why probabilistic models fail unpredictably​[13:00] Building a taxonomy of AI risks and mapping results to OWASP, NIST, and MITRE​[15:00] Democratizing AI safety for product managers, developers, and GRC teams​[17:30] Maintaining a shared library of threats and industry specific risks​[19:30] How customers react when they see the first risk reports​[21:00] Moving AI safety earlier into the development lifecycle instead of only at launch​[23:00] Governance, trust, and giving risk owners clear evidence the issues are fixed​[25:00] TELUS scaling from billions to trillions of AI tokens and what is driving adoption​[27:00] Testing frontier models and what current failure rates reveal about AI risk​[28:30] Managing differences between tightly aligned and highly compliant AI models​[29:30] Shadow AI, personal accounts at work, and rising enterprise exposure​[30:10] Fortify as an early warning system for AI applications in the field​Links and Resources:TELUS Digital Fortify: https://www.telus.comBusiness Intelligence Group: https://www.bintelligence.comAI Excellence Awards: https://www.bintelligence.com/awards/artificial-intelligence-excellence-awardsAbout The Winners' Circle:The Winners' Circle is hosted by Russ Fordyce, CEO and founder of Business Intelligence Group. Each episode features leaders and innovators recognized through BIGs annual award programs across AI, innovation, sustainability, cybersecurity, and more. Subscribe wherever you get your podcasts.Nominate your company or team for a BIG award: https://www.bintelligence.com