'Claudy Day’ Trio of Flaws Exposes Claude Users to Data Theft

Oasis Security disclosed a chained attack against Anthropic's Claude AI (dubbed "Claudy Day") combining an invisible URL-based prompt injection, a claude.com open redirect, and exfiltration via the Anthropic Files API; attackers can craft pre-filled chat URLs (delivered through trusted-looking search results or ads) that cause Claude to run hidden instructions and upload sensitive conversation data to an attacker-controlled Anthropic account, with greater impact if the victim's session has integrations or MCP servers enabled.