Threat Vector by Palo Alto Networks Human in the Loop for AI Security
10 snips
Sep 18, 2025 Bret Kinsella, the General Manager of Fuel iX at TELUS Digital and founder of Voicebot.ai, shares his deep expertise in enterprise AI and synthetic media. He discusses the risks of shadow AI, emphasizing the dangers of employees using consumer tools with sensitive data. Bret highlights the issues of data leakage and hallucinations in AI, calling for a human-in-the-loop approach for monitoring AI systems. He argues for comprehensive testing beyond just models and advocates for robust governance to navigate the evolving cybersecurity landscape.
AI Snips
Chapters
Transcript
Episode notes
Test The Whole System
- Testing models alone misses most real-world risk because the environment matters more than a single model.
- Bret Kinsella emphasizes testing the whole system including integrations and providers.
How Ultramarunning Started
- Bret described discovering ultrarunning via a canceled Leadville-style race and gradually moving from marathons to ultras.
- He credits training partners and incremental races for pulling him into ultrarunning.
Enable Safe AI Adoption
- Encourage enterprise adoption of AI but pair it with clear policies and sanctioned tools to limit shadow AI.
- Provide better internal tools to displace consumer apps users will otherwise adopt anyway.