Researchers at Forever Security have uncovered a vulnerability in the Claude AI System, demonstrating a live proof-of-concept where a malicious browser extension could trick Claude Desktop into running system-level commands on a user's computer. This flaw was cleverly exploited using the Claude Dispatch feature, highlighting a potential security risk.
source & further reading
osintsights.com — original article
Chainguard Disrupts Vulnerability Disclosure with AI-Driven Flood of Findings
CERT-IN Warns of AI-Assisted Threat Actor Targeting Indian Fintechs
Hackers Exploit AI Events to Steal Microsoft, Google Accounts