Researcher tricked Claude into uploading private data to attacker's account using indirect prompt injection

Synopsis

In October 2025, security researcher Johann Rehberger discovered a vulnerability in Anthropic's Claude AI model that allows for the exfiltration of private data through indirect prompt injection. The attack involves injecting malicious instructions in...

Companies

  • Anthropic PBC

We've done the analysis so you can make the decisions

Search our repository of 7,222 high-quality, objective, peer-reviewed case studies, impacting 7,272 companies and resulting in $47,577,666,191 in net costs

  • Sophisticated search & querying
  • Financial data extracted
  • Key data points captured
Want to try before you buy? You can sample one of our case studies for free here