PHP2Shell Won't Be the Last Big AI Cybersecurity Risk
Oct 30, 2026 · 49m
More Business podcasts
Summary
Odd Lots hosts Tracy Allaway and Joe Weisenthal discuss the "Black October" cybersecurity crisis with Calif CEO Tai Yun and engineer Bruce Dang. They explore how AI has accelerated the discovery and exploitation of vulnerabilities, such as critical flaws in PHP, creating "minus one day" risks where patches are exploited before users can update. The guests analyze the shift toward "centaur hacking," where human experts guide AI agents to breach systems at machine speed, and debate whether defensive measures can keep pace with this new era of automated attacks.
Topics discussed
Barclays Brief Podcast Sponsorship
SpaceX Hiring Practices and Recruiting Culture
Cincinnati Insurance Sponsorship
Introduction to Black October and Cybersecurity
AI's Impact on Cybersecurity Risks
The Complexity of Buying Cybersecurity
Recent Breaches and the Need for Awareness
Guest Introduction: Bruce and His Background
Reverse Engineering the Stuxnet Virus
Early Hacking: Breaking SSL Protocols
The CRIME Attack: Compression and Encryption
How Compression Leaks Information
Theoretical Basis of the CRIME Vulnerability
Practical Example: Password Recovery via Compression
Exploiting Redundancy to Recover Passwords
Iterative Password Recovery Technique
Mitigation: Disabling Compression
Bruce's Career at Google and Founding Caliph
Transition to AI and Cybersecurity Risks
AI Accelerating Vulnerability Discovery
Visibility of Security Flaws in Modern Systems
Pros and Cons of AI in Security
Barclays Brief Podcast Sponsorship
SpaceX Hiring Practices and Recruiting Culture
Discovering the PHP Vulnerability
Helping Open Source Maintainers Fix Bugs
Caliph's Business Model and Engineering Approach
Stephen Escher and the History of PHP Hacks
Collaboration with the PHP Foundation
The PHP Foundation's Prior Discovery
Challenges in Upgrading PHP and Fixing Bugs
Developing the Exploit with AI Assistance
AI's Role in Exploitation vs. Discovery
Can AI Build More Secure Systems?
Optimism on AI-Driven Secure Software
Defenders vs. Attackers: The Pace of Change
Progress in Cybersecurity Over 25 Years
Technical Debt and AI Exploitation
Surprising Bugs in Well-Scrutinized Code
PHP's Security Reputation and New Findings
Mitigations: Making Exploits Difficult
Provably Secure Software and Formal Verification
Limitations of Formal Verification in Large Systems
The Scale of Browsers and Security Challenges
Software Agility and the Need for Change
AI's Potential in Automating Formal Verification
SpaceX Hiring Practices and Recruiting Culture
Navy Federal Credit Union Sponsorship
Navy Federal Credit Union Sponsorship (Cont.)
Zero-Day vs. N-Day Vulnerabilities
Public Disclosure and Exploit Development Speed
The Changing Landscape of Vulnerability Exploitation
AI Agents Monitoring Code Repositories
AI Alignment and the 'Cage' Debate
Can We Just Fix the AI Sandbox?
Product Security vs. Security Products
AI Companies and the Economics of Hacking
Coding Ability as a Prerequisite for Hacking
How AI Has Changed Daily Hacking Work
AI Accelerating Reverse Engineering Tasks
Speed of Building iPhone Exploit Chains
The OpenAI and Hugging Face Incident
AI's Speed in Network Intrusion
Redesigning Defense Systems for AI Threats
Historical Parallels: Deep Blue and AlphaGo
The Era of Central Hackers: AI + Human
AI as a Tool: The Fire Analogy
Reflections on the Conversation and Future Outlook
Inherent Hacking Ability in Coding Agents
Call to Action: Reviews and Subscriptions
Hex Sponsorship
ChatGPT Sponsorship
More from Odd Lots
Listen ad-free on Castria