In this episode of Decoder, host Nilay Patel speaks with Alex Stamos, Chief Security Officer at Cognition, about how AI has fundamentally broken the cybersecurity landscape. Stamos argues that AI-powered tools have dramatically lowered the cost of attacks, enabling script kiddies and nation-states to breach systems faster than ever, leading to a 'dark age' of security. He advocates for proactive defenses, including rewriting critical infrastructure like the Linux kernel in memory-safe languages and accepting new risks like AI red-teaming.
Summarized by Podsumo
AI has inverted the cyber kill chain, allowing attackers to find and exploit vulnerabilities in real-time with tools like RTeX, making the cost per attack near zero.
The economics of defense are broken: attackers' marginal costs are droping while defenders' costs remain high, necessitating cheaper defense solutions.
Stamos predicts a 'dark age' of cybersecurity, but believes the industry can shorten it by rewriting legacy code (e.g., Linux kernel in Rust) and automating bug fixing with AI.
Companies must start red-teaming themselves with AI tools immediately to preempt breaches, despite legal and risk-management hurdles.
The era of relying on patching is over; exploitation timelines are too fast, and fundamental infrastructure like the Linux kernel needs to be rebuilt from scratch.
Stamos argues for verifying real content cryptographically rather than trying to detect AI-generated fakes, which is a losing battle.
"_'We are entering a dark age... the last time it felt like this was when I was a much younger man and I would go to the very early DEF CONs.'_ — Alex Stamos"
"_'The cost per token for attackers is going towards zero, and the cost for defenders is staying quite high.'_ — Alex Stamos"
"_'You rather have your own team break your system and fix it, or a 19-year-old in China do it for you.'_ — Alex Stamos"