Always-on AI agents that run on their own cloud computer are rolling out in ChatGPT. In organisations, they can be issued their own identity and credentials.
OpenAI announced the agents, called dots, on 29 September. In its launch post, the company describes them as "remarkably capable, always-on agents built to handle everything".
Dots are powered by the company's GPT-6 Astra model. Each has its own cloud computer and browser, learns from feedback over time and can work towards a user's goals around the clock.
Through OpenAI's plugin ecosystem, dots can connect to more than 4,000 applications. Users can reach them through ChatGPT, Slack or Microsoft Teams, and by voice call.
A dot can run several projects at once. OpenAI says users can open the dot's computer at any time to inspect its work. They can also let it connect to their own laptop.
Specialist dots for organisations
The governance implications are greatest for specialist dots, which take on dedicated responsibilities inside an organisation. OpenAI is previewing them through focused enterprise pilots.
The company says each specialist dot gets its own identity for access management. It also has its own credentials and access to the systems it needs. Support for IT-provisioned hardware and deep integration with systems of record is also planned.
OpenAI says it has tested specialist dots internally across procurement, invoice processing, email marketing, customer support and commercial contracting. Its engineering teams will work with pilot organisations to define each dot's responsibilities, tools and review processes.
OpenAI is also working with Microsoft to integrate specialist dots with the enterprise governance and security controls in Microsoft Agent 365. The aim is to let businesses manage dots through Microsoft tools they already use.
Controls and approvals
Users choose which applications a dot can access, using existing ChatGPT app controls. Dots start with built-in rules on when to act independently and when to seek approval.
Custom Rules let users allow specific actions, require approval or block them entirely. Built-in safety requirements always apply, and progress, including background work, is visible in an Activity View.
Actions that could affect accounts or share information are checked by an auto-review process against the user's instructions, rules and safety requirements. Certain sensitive tasks, such as changing a password, always stay with the user.
When not actively working on a task, a dot carries out what OpenAI calls "proactive research". It does this through connected applications using tools restricted to read-only access, so it cannot send messages or change content.
Dots can sign in to supported websites with saved passwords without exposing them to the model. OpenAI says monitoring systems can pause or stop a dot's work if a safety concern is detected.
Data use and availability
OpenAI says it does not use content from ChatGPT Business, Enterprise or Edu workspaces to improve its models by default. On personal plans, users can control whether dots' conversations and work are used for training.
The company says it does not train directly on proactive research or on a dot's notes to itself. It warns that "dots can still make mistakes, so always review consequential work".
Dots are rolling out to Pro and Business Premium users in eligible markets, with the first dot included at no extra cost. Enterprise users, including Edu and Healthcare, can try a beta once a workspace administrator enables it.
OpenAI says users will later be able to add more dots and scale the speed or volume of work each one takes on. It envisages teams of dots working together.