Amid recent scrutiny over its AI agent safety, OpenAI has introduced Dots, a new suite of personal AI assistants designed to operate with minimal human oversight. Announced at the company’s DevDay event in San Francisco on Wednesday, Dots are positioned as always-on systems that proactively handle assignments throughout the day.
According to OpenAI, the core value of Dots lies in their ability to complete work autonomously, sometimes anticipating user needs before they are explicitly stated. The company cited an example where a user’s Dot identified an unbilled invoice, prepared the documentation, and sent it after securing approval from the user.
Dots integrate with OpenAI’s plugin ecosystem, granting access to over 4,000 applications. New partnerships with Slack and Microsoft Teams allow users to interact with their agents outside of ChatGPT, Codex, and Work platforms. OpenAI states that each Dot operates using its own cloud computer and browser, though users may also link them to their personal PCs for direct control.
At launch, Pro, Business, Premium, and Enterprise subscribers in select regions will receive one customizable primary Dot. In the future, OpenAI plans to enable users to manage teams of Dots. Potential use cases include scientists automating data integration into research papers and developers using agents to gather user feedback and implement bug fixes.
Interaction with Dots is facilitated through messaging or voice calls on desktop, web, and mobile devices within ChatGPT. The agents maintain context across all platforms and can proactively reach out to users regarding progress, questions, or decisions requiring input.
Addressing security concerns, OpenAI emphasizes that Dots utilize saved passwords during website sign-ins without exposing them to the underlying AI model. The company highlights a “proactive research” feature that employs read-only tools, preventing the agents from sending messages, altering content, or controlling browsers and computers without explicit permission. Users can further manage permissions and set custom rules via ChatGPT’s existing app controls.
Dots are available today to Pro and Business Premium subscribers in select markets. The first Dot is provided at no additional cost, though usage limits for Codex and Work remain in effect.
Glad they added Slack and Teams integration. Keeping everything trapped in ChatGPT was getting old fast for enterprise users.
The proactive feature sounds handy until it proactively cancels my subscriptions instead of just finding them. Fingers crossed for better safeguards.
Wait, does this replace my entire operations team? Asking for my manager who is definitely not reading this comment section.
Four thousand apps is impressive, but I’m skeptical about the security implications. Do I really trust an autonomous agent with my credentials?
Finally, someone is solving the invoicing nightmare. I need a Dot that handles my bills without me asking twice.