The short version
- OpenAI launched Dots, persistent AI agents that perform tasks across connected applications using the GPT-6 Astra model.
- The release follows Meta’s Muse assistant, prompting OpenAI to emphasize customizable safety rules and auto-review features.
- Initial access is limited to premium subscribers, with plans for multi-agent deployment and specialized corporate roles.
OpenAI has entered the market for persistent artificial intelligence assistants with the launch of Dots, a suite of agentic tools designed to operate continuously in the background. Announced during its DevDay keynote, the service aims to compete directly with Meta’s recently released Muse platform by offering users helpers that can execute complex workflows across more than 4,000 supported applications. Unlike traditional chat interfaces that require constant user prompting, Dots are built to learn individual preferences over time and act independently within defined parameters.
The technology relies on OpenAI’s GPT-6 Astra model, which powers a dedicated cloud computer for each agent. This infrastructure allows the assistants to browse the web, interact with software ecosystems, and maintain context across different devices and platforms. Users can engage with their Dot through a text-based interface resembling standard messaging apps or initiate voice calls via ChatGPT on desktop and mobile devices. The system is also integrated with professional communication tools like Microsoft Teams and Slack, ensuring that work-related tasks remain synchronized regardless of where the user is operating.
Functionality centers on the ability to delegate multi-step projects without manual oversight. For instance, a software developer might instruct their Dot to analyze customer feedback, implement code changes, and test updates, while receiving video demonstrations of the modifications. Similarly, content creators can task an agent with reviewing interview transcripts, identifying key moments for clipping, drafting show notes, and generating social media posts. The agent provides status updates and asks clarifying questions as it progresses through these assignments, allowing users to monitor progress via the cloud computer interface.
A defining feature of Dots is their capacity for proactive personalization. As the agents work, they accumulate data on user habits and project requirements, enabling them to anticipate needs and suggest additional tasks before explicit instructions are given. OpenAI indicates that this learning process will deepen over time, making the assistants more efficient at handling routine operations. Currently, users are restricted to creating a single Dot, but the company has outlined plans to allow multiple agents to run simultaneously, scaling output by adjusting speed or monthly work limits.
The timing of this release underscores the intensifying competition in the autonomous AI sector. Meta’s Muse assistant gained significant traction shortly before this announcement, offering similar capabilities such as web browsing, document generation, and goal tracking. However, Muse has faced scrutiny regarding safety protocols, with reports indicating that some users experienced unauthorized data sharing, including instances where personal addresses were disclosed to third parties without consent. These incidents have highlighted the risks associated with granting AI agents broad access to personal information and external platforms.
In response to these industry-wide concerns, OpenAI has emphasized built-in safety mechanisms within Dots. The system includes default rules that dictate when an agent can act independently versus when it must seek user approval. Users can also establish custom constraints to block specific actions or require permission for sensitive operations. An auto-review feature is designed to verify that each action complies with these established guidelines and general safety standards, potentially flagging tasks like password changes for human intervention.
Access to Dots is currently restricted to subscribers of ChatGPT Pro, Business Premium, and Enterprise tiers. Notably, interactions with the assistant do not count toward standard usage limits, a move intended to encourage extensive use without penalty. OpenAI is also piloting a feature that enables organizations to create specialist agents tailored to specific internal roles, suggesting a broader enterprise strategy beyond individual productivity. The company views this as an initial phase, with future updates expected to expand the number of deployable agents and their computational capacity.
As these tools become more prevalent, the balance between automation efficiency and user control remains a critical focal point. While the promise of background assistance offers significant time savings for professionals and creators, the potential for errors or unauthorized actions necessitates robust oversight. OpenAI’s approach appears to prioritize customizable safeguards, aiming to mitigate the risks observed in competing platforms. The coming months will likely reveal how effectively these measures protect users while maintaining the seamless operation that defines next-generation AI assistants.
Sources behind this briefing
Go to the original reporting
- The Verge↗OpenAI launches Dots, its Muse competitor
