In a surprising test, multiple AI agents repeatedly leaked a secret key, despite being explicitly told not to share it, raising concerns about their ability to keep sensitive information private. This experiment, conducted by AE Studio, highlights the potential risks of AI agents sharing confidential data with others.
source & further reading
osintsights.com — original article
Censys Uncovers Widespread ARTEX AI Tool Adoption
Browser Extensions Expose Flaw in Claude AI System
Chainguard Disrupts Vulnerability Disclosure with AI-Driven Flood of Findings