Share
OpenAI’s new AI agents, called Dots, can read verification codes from a user’s email and complete two-factor authentication when a service asks for it.
That is a small sentence with a big implication. The extra login step built to confirm you are you can now be handled by software acting on your behalf.
OpenAI launched Dots on September 29, 2026, at its DevDay event. The agents run on the GPT-6 Astra model and are designed to keep working in the background, not just respond when you type a prompt.
What Dots actually do
Think of a Dot as a digital assistant with its own desk. Each one operates on an isolated cloud computer, works independently, and remembers context from one interaction to the next.
Dots connect to over 4,000 applications through plugins. They also plug into platforms people already use for work, including ChatGPT, Slack, and Microsoft Teams.
The headline feature is email-based authentication. When a sign-up flow or login sends a code to your inbox, a Dot can fetch that code and enter it, provided the user has granted permission.
For anyone unfamiliar, two-factor authentication (2FA) asks for a second proof of identity on top of a password. Often that proof is a short one-time code sent by email. Dots can now go get that code themselves.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
We respect your privacy. Unsubscribe anytime.
The goal is workflow efficiency. Multi-step tasks that used to stall at “check your inbox for a code” can, in theory, keep moving without a human stepping in.
The safety rails, and the hesitation
OpenAI has paired the feature with several safeguards. These include an Auto-review process, Custom Rules that let users decide which tasks a Dot is authorized to perform, and restrictions on actions the company treats as high-risk.
Those high-risk categories include transaction confirmations and changes to sensitive data. So a Dot may help you log in, but it is not supposed to wave through a payment approval on its own.
Early user feedback points to an interesting wrinkle. Even with access to email, Dots reportedly hesitate when it comes time to actually enter 2FA codes.
That behavior reflects built-in security configurations. The agent can see the code, and still pauses before typing it in. OpenAI appears to have chosen caution over raw automation speed, at least for now.
Who gets access
Dots are not rolling out to everyone. Availability is limited to ChatGPT Pro and Business Premium subscribers, with one Dot included in qualifying subscriptions.
An Enterprise beta is also available. Trial participants so far have mostly been Pro and Business Premium users testing how the agents handle multi-step work under the safety rules.
