ChatGPT Dots Turn On by Default Nov 18: What Admins Must Know
OpenAI flips the switch on always-on ChatGPT agents for Enterprise users on November 18, and billing starts the exact same day.
Tools ยท Source: OpenAI
What happened
On November 18, 2026, OpenAI will make dots available by default to every member of eligible ChatGPT Enterprise, Edu, and Healthcare workspaces. A dot is an always-on agent that runs on its own cloud computer. It keeps working across tools and projects even when a user turns off their laptop. People can reach their dot from ChatGPT, Slack, Microsoft Teams, texting, or phone calls.
The free admin preview ends on this exact date. If administrators do nothing, the workspace default flips to allow access, and billing for enabled dots begins immediately. Workspaces with UAE data or inference residency are excluded from this rollout. Members still have to actively create their personal dot, meaning nobody gets an agent automatically, but the permission to build one will be universally granted unless admins intervene.
OpenAI plans to offer a flat-rate billing option based on the type of work a dot does, alongside a standard consumption-based model. However, the billing structure contains a massive catch. Tasks that a dot delegates to other OpenAI products like Work or Codex will be billed separately on their own meters.
Key facts
- November 18, 2026 โ Date dots become default-on and billing begins
- 12 โ Days admins have to opt out or limit access from the report date
- UAE โ Workspaces excluded based on data or inference residency
Why it matters
This changes how teams integrate AI into their daily operations. Builders and engineers get access to an autonomous coworker that can handle background tasks, run parallel processes, and operate on recurring schedules. You no longer have to wait for a prompt response. The agent works while you sleep. This shifts the AI paradigm from a passive chatbot to an active participant in your engineering and business workflows.
The second-order effect hits corporate security and finance hard. Dots use a cloud browser that holds separate sessions from an employee device. Code running on the dot cloud computer can potentially read app tokens or browser cookies. Finance teams also face unpredictable costs. A single dot can spin up multiple Codex and Work tasks, each adding to the monthly bill. There is no documented dot-specific spend cap in the admin controls.
For builders
Audit cloud browser session security
Dots hold their own browser sessions in the cloud. Security teams lose visibility and control if they do not configure password manager settings to match company policies.
Prepare for unpredictable compute costs
A dot can delegate work to background agents that run in parallel. Finance teams pay standard rates for every Codex or Work task the dot triggers, with no hard cap documented.
Scope pilot access with custom roles
Admins can limit exposure by turning off the workspace default. Assign a custom role to a small pilot group before November 18 to test capabilities safely.
My take
OpenAI is forcing enterprise adoption by making admins opt out instead of opt in. Launching an autonomous agent that can rack up uncapped Codex bills while holding active session cookies in the cloud is a massive security headache. I tell every founder I know to ask their IT team what their dot can touch before they wake up to a compromised system or a blown budget.
Original reporting: OpenAI. This is my rewrite and opinion.