CrewAI Hit by Critical Vulnerabilities Enabling Sandbox Escape and Host Compromise

CrewAI is reported to contain four critical vulnerabilities in its Code Interpreter and related tools that permit remote code execution, SSRF, and arbitrary local file reads by exploiting insecure fallback behaviors (e.g., fallback to a SandboxPython environment when Docker is unavailable) and improper input validation. The vendor has acknowledged the issues and plans mitigations; administrators are advised to disable the Code Interpreter, set allow_code_execution=False, sanitize untrusted inputs, and monitor Docker availability until patches are released.