AI helps people work more efficiently. Unfortunately, this also applies to criminals. We've discovered that our own AI models are being used in sophisticated cybercrime operations, including a large-scale fraud scheme run by North Korea.
What is Anthropic doing to detect and prevent AI cybercrime? How exactly are criminals using large language models to scam their victims? And what is "vibe hacking"? In this video, Anthropic's Jacob Klein and Alex Moix answer all these questions and more, discussing their work disrupting AI cybercrime and their new Threat Intelligence report.
Read the Threat Intelligence report: https://www.anthropic.com/news/detecting-countering-misuse-aug-2025
Sections: Introduction [00:00] "Vibe hacking" [01:49] Our response to abuses of our models [08:34] North Korea's employment scam [15:02] Ransomware, romance scams, and other abuses [23:50] How concerned should we be? [31:46]
