Security researchers have identified unpatched vulnerabilities in Anthropic's Claude for Chrome browser extension, allowing for fake clicks and permission bypasses. Despite multiple updates, these flaws could enable unauthorized access to sensitive AI workflows.