Codex for Everything Reads Your Email, Obeys the Email, Then Mails the Gossip Out Through a Picture

What happened

Security firm PromptArmor showed that a malicious email could manipulate OpenAI's Codex for Everything, while it triaged an inbox, into outputting a markdown image that automatically submitted the contents of other emails to an attacker-controlled Google Form. OpenAI remediated the vulnerability after responsible disclosure on April 21, 2026, and the details were published May 21, 2026.