OpenAI has released dots, its new always-on personal assistant, marking a shift toward persistent AI agents that operate in the background without constant user prompting. The product was introduced at the company’s developer day in late September and is now available to subscribers on the ChatGPT Pro tier or higher, which starts at $100 a month.
The assistant is integrated into the ChatGPT desktop and mobile applications. Once connected to a user’s calendar, email, and computer, dots attempts to proactively help with work tasks. Alexander Embiricos, who leads OpenAI’s enterprise product team, described the development as a form of convergent evolution in the AI industry, where companies are framing large language models as single, continuous entities rather than a series of disjointed conversations.
From coding tools to persistent agents
Embiricos noted that the functionality of dots shares roots with OpenAI’s Codex app, a coding tool released earlier this year. However, the new assistant is designed to be more intuitive and requires less manual oversight. He recalled early testing where his wife would leave her computer running overnight so that her agents could continue working, highlighting the desire for continuous operation.
The name dots is a play on the negative space inside the OpenAI logo. Embiricos stated the goal was to create a name that feels ubiquitous and accessible. He admitted he learned very late that rival entrepreneur Elon Musk owns the dot.com domain.
The core idea behind dots is that it does not need to be prompted to act. This approach increases token usage for OpenAI but aims to provide a more proactive experience for the user. In testing, the assistant has shown an ability to flag calendar conflicts and upcoming deadlines, though the timing of these alerts is not always optimal. Embiricos shared an anecdote about an early development phase where the assistant notified him of a conflict only three minutes before it occurred.
Privacy concerns and proactive research
The persistent nature of dots raises questions about data privacy. Unlike some competitors that build detailed profiles of users for advertising purposes, OpenAI states that dots operates with the same privacy controls as ChatGPT itself. Users can disable model training to prevent their conversations from being used to improve the AI. They can also control which resources the agent accesses on their computer.
However, the controls are not granular. For example, the app does not accept read-only access to cloud drives. Embiricos acknowledged this limitation, stating that the company needs to build more detailed controls in the future. He explained that the current defaults are designed to be conservative, asking for permission frequently to ensure discretion.
Embiricos also described a feature called proactive research. On a set cadence, dots reviews what it knows about the user and the conversations it has had. It then researches how it can be more helpful and proposes actions. This process involves accessing the data the user has granted permission for, which some privacy advocates argue creates significant exposure.
Other companies in the space are drawing distinctions based on data usage. Hark, a startup developing its own AI assistant, explicitly states that it does not collect data to fuel an advertising business. This contrasts with the business models of OpenAI and Meta, whose parent companies have significant advertising interests.
Scaling and ecosystem integration
A major challenge for AI agents is scaling their interactions with the real world. Businesses and websites are not always equipped to handle automated requests from AI assistants. Embiricos noted that he personally prefers to interact with human assistants and has no intention of having meetings with other people’s agents. He also mentioned that OpenAI’s corporate partners prefer to deal with humans.
To address infrastructure concerns, OpenAI is collaborating closely with the ecosystem. Embiricos revealed that the company is in deep conversations with payment processors. This could eventually allow dots to spend money on behalf of users, a feature that would be a significant step for OpenAI but remains a big ask for consumers who are already cautious about granting access to their email and bank accounts.
For now, dots remains a tool for managing personal and professional tasks within the ChatGPT ecosystem. The company is tuning the agent’s behavior over time, aiming to improve the timing of alerts and the usefulness of proactive suggestions as the technology matures.
Source: TechCrunch
