Is It Time To Rebuild the Internet?
Oct 30, 2026 · 43m
More Technology podcasts
Summary
Host Charlie Warzel and guest Alex Stamos, CISO at Cognition, discuss the "Black October" cybersecurity crisis, where AI agents have automated the discovery and exploitation of vulnerabilities in critical infrastructure like PHP and Node.js. Stamos argues that the current reactive approach is unsustainable and advocates for proactively rewriting legacy code in safer languages like Rust to shorten the "dark age" of digital insecurity.
Topics discussed
Sponsor: InPress news app
Shortening the cybersecurity dark age
Fragility of digital infrastructure and legacy code
AI agents as relentless hacker armies
Impact on supply chains and healthcare
Introduction of guest Alex Stamos
What has changed in the last two weeks
Automation of kill chains via open-weight models
History of vulnerabilities in C and C++
Waves of exploitation and manual vs automatic attacks
GMO Internet hack and public panic
Internet infrastructure holding up despite bugs
The lag between finding and fixing bugs
Attackers getting 'promoted' in skill level
Loss of premier status for state actors
PHP vulnerability found by AI prompt
Coordinated disclosure challenges in AI era
Asymptotic approach to bug discovery
Formal verification and rewriting code with LLMs
The 'Dark Age' of cybersecurity metaphor
Proactive vs reactive security strategies
Rebuilding security teams and structures
Sponsor: InPress news app
Risk of new vulnerabilities in rewritten code
Why C and C++ are malpractice in 2027
Low-level language errors and memory management
Advantages of modern languages like Rust
Impossibility of rewriting all legacy systems
Cost and societal value of infrastructure upgrades
Vulnerabilities in power grid systems
Economic incentives for grid security
Coordinated action issues in critical infrastructure
Limits of physical isolation for power grids
Security concerns for community hospitals
Recent attacks: Voyager and AI-enabled capabilities
Chinese actors targeting US AI ecosystem
Stealing API keys and model weights
Ecosystem of hackers for Chinese AI companies
Sponsor: InPress news app
Pace of innovation in defensive AI
Concerns about speed of AI deployment
Red teaming protections with cyber models
Human misuse vs model autonomy
Liability and defensive cyber capabilities
Logic of continuing AI development despite risks
Coordination problems in safety regimes
Lowering cost of defensive capabilities
Resiliency of systems and PHP bug aftermath
Coordinated disclosure as a safety net
Worry about custom AI agent attacks
Advice for normal people: password hygiene
Paranoia regarding AI voice and data theft
Using limited computers for safety
Secondary sock puppet accounts
Closing remarks and thanks
Galaxy Brain AMA promotion
Sponsor: InPress news app
More from Galaxy Brain
Listen ad-free on Castria