OpenAI's new assistant, dots, keeps working even when you are not asking

OpenAI has launched dots, an AI assistant that keeps working in the background even when you are not asking it for anything. It debuted at the company's developer day in late September. In TechCrunch's testing, it caught calendar conflicts and upcoming deadlines, though not always in time to act on them.

What dots does

Most chatbots wait for a question. Dots is built to act first. OpenAI calls one feature proactive research. On a regular schedule, dots looks at what it knows about you, your past conversations and what you have given it access to. Then it suggests ways to help.

OpenAI treats dots as one entity in a single ongoing conversation, not many separate chats. Alexander Embiricos, who leads OpenAI's enterprise product team, said that framing makes it feel intuitive. The name plays on the empty space inside OpenAI's logo.

How it performed

The early results are mixed. Embiricos said the first conflict his own dots flagged came three minutes before the event. He expects timing to improve. He also described an engineer at a drone delivery startup whose dots spotted an outage, studied it and came back with ideas for a fix before the engineer noticed the problem.

Privacy and access

An assistant that watches your files raises an obvious question: how much does it see? OpenAI says dots has the same privacy controls as ChatGPT. The controls are limited, as the list shows.

• You can turn off model training, which stops OpenAI from using your conversations to improve its AI.
• You can control which resources dots reaches on your computer, and how.
• The controls are not granular, and dots refuses read-only access to cloud drives.

Embiricos said OpenAI needs to build "super, super, super granular controls." He also said early versions asked permission so often that users complained. Meta's rival assistant, Muse, has drawn criticism for the profiles it builds of users' family and friends. A smaller startup, Hark, says it does not collect data to fuel advertising.

What is not known yet

The material does not say what dots costs or who can use it today. It also does not say when the timing and permission settings will improve. Embiricos himself says he never wants a meeting with someone else's AI agent, and he still keeps a human assistant because OpenAI's partners want to deal with a person.