Firewalls Don't Stop Dragons Podcast Here’s to 500!!
21 snips
Sep 28, 2026 
Bruce Schneier, renowned security technologist and author, returns for a milestone conversation on AI’s next chapter. They explore how autonomous agents work, prompt injection, identity and liability, the accelerating security arms race, privacy and portable data, mass surveillance, and the case for public AI infrastructure. Plus, a Captain Crunch whistle story, anniversary surprises, and a look back at 500 weeks of security education.
AI Snips
Chapters
Books
Transcript
Episode notes
Agents Need Owners And Separate Identities
- Agents should have identifiable owners, separate credentials, limited delegation, and auditable responsibility rather than impersonating users.
- Schneier compares an agent to a dog: society needs to know whose agent caused the damage.
AI Liability Should Follow Existing Legal Principles
- AI liability can involve users, deployers, model makers, and employees, but courts already apportion fault in complex accidents.
- Schneier argues that corporations should face penalties when autonomous systems commit acts that would send humans to jail.
AI May Give Cyber Defense The Upper Hand
- AI will accelerate both vulnerability discovery and patching, potentially making defense stronger when fixes happen before deployment.
- Shared vulnerability intelligence favors defenders, while attackers gain advantage if they coordinate more effectively.
