ChatGPT's Dots have picked up a practical habit: when a site sends a sign-in code to an inbox you have already connected, your Dot can fetch that code and use it to log in on your behalf. You are asked first — allow it once, or always let it fill the code in.

OpenAI's Dots lead, Samir, quoted the upgrade note and wrote that it had "just rolled out," with more Dot improvements and fixes inbound. The official ChatGPT account reposted the same upgrade, and reporting on November 7 describes the agents reading verification codes from email and completing two-factor sign-ins when a user has granted permission.

Early user feedback adds a wrinkle: even with inbox access, Dots reportedly hesitate when it comes time to actually enter a code, which reporting attributes to built-in security settings. The posts do not list which inboxes or regions are included.