Shadow AI Epidemic: Uncovering Agents on the Endpoint, British Library Breach, & News - Amit Assaraf - ESW #475
Summary
Episode 475 of Enterprise Security Weekly features an interview with Amit Asaroff of Palo Alto Networks on securing AI agents and non-binary software on endpoints using the Koi platform. The hosts also analyze the British Library’s prolonged recovery from a ransomware attack, highlighting the dangers of extreme technical debt and the difficulty of replacing legacy systems. Additional segments cover recent enterprise security news, including funding trends, ID leaks, and a late-breaking BGP incident.
Topics discussed
Episode intro: AI agents, British Library, and news overview
Casual chat: Labor Day, beer, and tea preferences
Promo: Attack Surface Management Virtual Summit
Interview start: Securing AI agents with Palo Alto Networks
Koi's origin: The Dracula experiment and non-binary security
The rise of non-binary threats and AI agent risks
Securing developer endpoints and AI agent visibility
Koi's mascot, culture, and the OpenClaw era
Guardrails for AI creativity and the lethal trifecta
Methodology: Data layer, categorization, and controls
Employee experience and communication in security
Customer stories: Balancing security and productivity
Human-in-the-loop approvals and YOLO adopters
Palo Alto acquisition strategy and future of computer use
Local AI models, Mac Minis, and zero trust ads
British Library attack: Tech debt and infrastructure loss
The cost of technical debt and organizational culture
Ransom decisions, credential stuffing, and MFA
Sponsor: Rubric Agent Cloud for agent failure
Vibe check: AI agents for defenders and triage bottlenecks
Funding: Penetration testing startups and market overlap
Acquisitions: ProofPoint, Veronus, and valuation trends
Shutdowns: Minimus and the zombie corn phenomenon
Vulnerability data: Exploit timelines and threat intelligence
Risk reduction: Prioritizing patches over 100% coverage
Leaked AWS keys and the need for expiration dates
BGP and DNS vulnerabilities: Infrastructure layer risks
AI jailbreaks and casual vulnerability discovery
Leaked IDs and facial recognition risks
Tech news: Robot ducks, AI toothbrushes, and IoT
Outro and next episode preview
Listen ad-free on Castria