Anthropic Patches Silent Zero-Click 'SneakPeek' Flaw in Claude Code Desktop

A high-severity vulnerability in Anthropic’s Claude Code Desktop allowed attackers to steal sensitive credentials without user interaction. The flaw, named ‘SneakPeek,’ exploited weaknesses in the preview pane’s architecture, bypassing sandbox protections. Initial fixes were ineffective, requiring deeper architectural changes. The vulnerability could exfiltrate AWS credentials, SSH private keys, and OAuth tokens silently. Researchers warn of similar stealth vulnerabilities in AI environments and advocate for robust security solutions.