The safe_eval function in Ansible before 1.5.4 does not properly restrict the code subset, which allows remote attackers to execute arbitrary code via crafted instructions.
https://www.securityfocus.com/bid/68232
https://github.com/ansible/ansible/blob/release1.5.5/CHANGELOG.md