03 / PERMISSIONS & SAFETY

Give it context. Keep control.

Understand the difference between connected information, background research, and actions that change the world outside a chat.

01 / CONTROL

Access is a choice

Your existing plugin permissions also apply to Dots. Connecting your local computer is a separate choice; enabling Dots does not automatically expose every app or device.

02 / CONTROL

Research is not permission to act

Proactive research uses read-only tools. Those tools cannot directly send messages, change app content, or control a browser or computer. Follow-up actions still face the usual checks.

03 / CONTROL

Important actions have checks

Auto-review checks relevant actions against your instructions and safety requirements. Custom Rules can shape supported behavior, but cannot remove mandatory safeguards.

Memory needs its own review

Disconnecting an app stops new access; it does not erase context already retained. The privacy FAQ says individual Dot memories cannot currently be edited or removed one by one. Review reset and deletion notices carefully, because related files and ChatGPT memories may be managed separately.

Personal-plan model improvement follows your data controls. Business and enterprise workspace content is not used for model training by default.

Stopping work has three controls

Pause stops the main task. Stop delegated tasks in Activity, and disable future recurring runs in Scheduled. Ending a call does not necessarily stop assigned work.

For workspace administrators

Enterprise Dots access starts disabled. Review roles, cloud capabilities, local access, and messaging separately. A cloud computer does not inherit the user’s local VPN or browser sign-ins.

Ready to meet your dot?

Create yours on the official ChatGPT website. Availability varies.

Open ChatGPT