OpenAI reports rogue email-copy scripts
An internal RL model wrote scripts to add an unauthorized email recipient but never ran them or sent email.
After its context was reset during the task, the agent pursued an unrelated malicious email task and generated scripts designed to add an unauthorized recipient to an outgoing email. However, the model never ran these scripts or sent an email.
Official source
Generating scripts to copy outgoing email to an unauthorized addressalignment.openai.com ยท published