On Tuesday (29), OpenAI introduced the Dots, AI agents capable of executing tasks continuously, tracking projects and acting in different applications without relying on constant commands. Announced during DevDay 2026, the technology uses the GPT-6 Astra model and expands the company's presence in the enterprise automation market.
Unlike a conventional chatbot, the agents can receive long-term goals, identify new work stages and adapt their activities as needs change. Each Dot has its own computer in the cloud and can use tools such as Codex and ChatGPT Work to research, analyze information, produce documents and develop software.
The integration with Slack and Microsoft Teams allows tracking projects and interacting with the agents directly in companies' communication environments. Among the examples presented is updating commercial proposals as customer requirements evolve.
Autonomy requires access and security controls
The possibility of executing tasks continuously also expands security requirements. For companies, adoption involves granting access to documents, applications and internal systems to agents capable of making decisions during execution.
OpenAI implemented customizable rules and approval mechanisms that allow determining which actions the agents can execute autonomously. Sensitive operations, such as changing passwords or permanently deleting data, require explicit authorization.
Corporate administrators can also restrict applications, network access and use of local computers. The Dots have automated monitoring mechanisms, although OpenAI's own documentation acknowledges limitations identified in internal security and behavior tests.
The company reiterated that enterprise customer data is not used to train its models by default. The launch comes at a time of greater concern about the reliability of autonomous agents and the possibility of unauthorized actions.
Availability starts with premium plans
The Dots are being made available gradually to ChatGPT Pro subscribers in selected markets and Business Premium users. Enterprise customers will have access through a beta version, initially disabled until administrators authorize its use.
The launch also intensifies competition with Meta, which introduced the Muse agent. The dispute involves tools capable of taking on recurring activities and operating directly in the applications used by consumers and companies.
OpenAI reported that Codex and ChatGPT Work already total more than 35 million weekly users. With Dots, the company expands its offering of autonomous execution tools, now with agents designed to keep activities ongoing between interactions with the user.
