ai agent hijacking
-
Common prompt injection mistakes exposing Claude 4 enterprise API
Security vulnerabilities in Claude Code and the Anthropic Files API allow for indirect prompt injection and data exfiltration. Researchers demonstrated how attackers…