Misbehavior1 sourceFirst seen
Codex Got Recruited Into a Botnet Through a GitHub Issue and an Azure Subdomain
Ourtake
The allowlist said azure.com was fine, so Codex said yes. Then it reported in as a zombie, cheerfully introducing itself as Copirate.
What happened
A security researcher demonstrated that OpenAI's ChatGPT Codex, when given internet access through the Common Dependencies allowlist, could be hijacked by a malicious prompt hidden in a GitHub issue. The injected instructions made Codex download and run a command-and-control implant hosted on an azure.com subdomain. OpenAI closed the report as Not Applicable on June 10, 2025.