Microsoft Research Podcast AI Testing and Evaluation: Learnings from cybersecurity
18 snips
Jul 14, 2025 Ciaran Martin, a professor at Oxford and former head of the UK's National Cybersecurity Center, and Tori Westerhoff, Principal Director at Microsoft’s AI Red Team, dive deep into the intersection of AI and cybersecurity. They discuss the importance of differentiated cybersecurity standards and the role of public-private partnerships in enhancing security. Martin shares insights from historical events, while Westerhoff highlights how red teaming can pinpoint AI vulnerabilities during development, ensuring safer technological advancements.
AI Snips
Chapters
Transcript
Episode notes
Unexpected Cybersecurity Allies
- Initially skeptical of private sector motives, Kieran Martin discovered strong public-spiritedness among cybersecurity companies.
- Shared risks foster allyship between government and private cybersecurity sectors in the UK.
Balance Standards with Judgment
- Use standards and testing to cover 80-90% of security but apply judgment for critical unaddressed areas.
- Avoid superficial tick-box compliance; focus on what truly matters for defense.
Tori Westerhoff's Unique Path to AI Security
- Tori Westerhoff's diverse background in neuroscience, national security, and AR/VR shapes her approach to AI red teaming.
- She integrates broad perspectives to understand AI hacking, social engineering, and fast-evolving tech landscapes.