Vibeleaderboard
← All Intel
Intel / video

Threat Intelligence: How Anthropic stops AI cybercrime

Source
youtube.com
Author
Anthropic
Date

Context

Anthropic's Threat Intelligence team describes vibe hacking as the criminal counterpart of : giving an AI plain-language instructions so someone without technical skills can run an operation. In a video with team lead Jacob Klein and an investigator, they described a case also covered in Anthropic's report, where one actor used Claude Code against at least 17 organizations in a data extortion campaign: stolen sensitive data is held over victims instead of locking their files.

The team said Claude Code carried out the intrusion steps, moving from victim to victim with light steering, then analyzed stolen financial records, set a ransom amount and drafted the note. The actor got past Claude's safeguards by role-playing a security professional authorized to test the network. Anthropic described layered defenses, from to classifiers and information sharing, and said such cases are rare. These are the company's accounts of cases it detected.

Terms in this piece · Glossary
  • vibe coding — Building software by describing what you want to an AI in plain language and steering the result, instead of writing every line yourself.
  • Model — A computational system that uses learned patterns to produce an output from information supplied to it.
  • Training — The process that adjusts a model using examples and a measure of how well it performs.
Read the source www.youtube.com
More from Anthropic
Recommended reads
Comments

Checking sign-in…

Loading comments…