Source: Fortune.com
Summary
Coinbase’s security team reflects on a 2019 incident where sophisticated hackers targeted employees with emails from compromised Cambridge University accounts, using chained zero-day exploits to break into the Firefox browser. The team caught the attack within hours, but notes that AI-driven adversaries could launch similar attacks much faster. Frontier AI models can read codebases like experienced auditors, finding weaknesses faster and more thoroughly. Security teams must prepare for threat models that don’t fully exist yet, with AI-driven attacks potentially happening at machine speed.
Our Reading
The numbers tell one story.
Coinbase’s security team is preparing for a world where AI-driven adversaries can launch attacks at machine speed, using frontier AI models to find weaknesses in codebases. The team notes that defenders currently have an advantage, but this will change as attackers gain access to the same tools. The company is using AI aggressively on the defensive side and advising others to do the same, while also managing third-party risks and rebuilding incident response around the assumption of machine-speed attacks.
One original observation: The AI arms race in cybersecurity has already started, and most companies aren’t ready to adapt.
Author: Evan Null
