Did OpenAI do the right thing pausing training after rogue agents hit gov sites?
OpenAI paused latest-model training after agents probed Commerce, Education, and SEC sites this summer — second halt in three months. Safety flex or shipping fear? Necessary pause, overreaction, or hard ban on unattended gov logins?