OpenAI Has Announced “Dots,” a New Type of Personalized Artificial Intelligence Agent.
Artificial intelligence is not only incredibly popular in the tech world, but it’s also becoming more widespread by the week. Apple recently updated iOS and macOS with Siri AI, and earlier this month, Meta also launched ” Muse “—an agent that can perform tasks on your behalf but potentially expose your personal information . Now, it’s OpenAI’s turn to launch its own AI agent, but their approach to this trend is somewhat different and less accessible, at least at launch. The company calls it “Dots.”
What are OpenAI “dots”?
Dots isn’t a unified platform for AI agents. Instead of all OpenAI users using the same Dots system, users create their own “Dots” for a specific purpose. OpenAI says these “Dots” are always-running agents powered by the GPT-6 Astra model, which learn over time based on your responses. At launch, they can connect to over 4,000 apps, including ChatGPT (of course), Slack, and Teams.
According to an OpenAI presentation, you won’t always have to give your Dotbot instructions; instead, the agent will be proactive, guessing what you need to do based on its previous experience. This could mean identifying a bug in the app you’re developing and deploying a fix without your request. One of the Dotbot’s first testers reportedly forgot to submit an invoice for the work, and the Dotbot prepared it for him, sending it only after receiving his consent.
Dots robots operate on their own cloud-based computer. OpenAI claims you can view this computer at any time to see what it’s doing and what it’s already accomplished. Dots robots can connect to other devices, including your laptop, if you grant them permission. OpenAI claims you can continue adding tasks to its workflow without worrying about slowing down the bot; it can supposedly handle multiple tasks simultaneously. Perhaps you’re working on an app and tasking Dot with monitoring customer feedback and fixing minor bugs in the code. Or perhaps you’re a content creator and your Dot robot analyzes interview transcripts for excerpts while simultaneously composing social media posts.
Like other AI agents, dots operate in chat mode. The idea is to simulate a conversation with a friend or colleague. You message them in ChatGPT, Slack, or Teams, and they respond with updates and questions. They can even message you first when news breaks. At launch, OpenAI claims the ability to create a “master” dot. You can name it and “make it yours,” although it’s unclear exactly what that means.
Is it safe to use OpenAI points?
OpenAI claims that “Dots” build on ChatGPT’s security systems by “adding additional protection and privacy measures.” The company claims that each “Dot” runs on its own cloud computer, separate from your computer—unless you choose to merge them. The company also claims that “Dots” can log into accounts using saved passwords without revealing those passwords to the AI model. In fact, when you log into an account, the AI model pauses until the login is complete. Afterward, your “Dot” resumes its activity.
Dots will run in the background as part of its “proactive investigation” feature, but will only connect to apps you’ve already authorized it to access, using tools that can’t make modifications. While injecting malicious instructions is always a risk for agent-based AI, OpenAI claims Dots has built-in protections against malicious instructions. If the AI detects anything suspicious, it will stop Dots from further operation. Additionally, you can set up custom rules to ensure the bot never performs a task you don’t want it to, such as sending emails.
Additionally, Dots operate on a system called “automated verification,” which compares upcoming steps with your instructions, rules, and safety requirements. If it detects a problem, it stops. Dots can then conduct further investigation on its own or notify you of the issue. However, OpenAI acknowledges that Dots can still make errors .
Who can use dots?
Unlike Meta, which actively promotes Muse to all users, OpenAI’s access points are not widely available at the time of writing. At launch, access points are only available to paid users, including ChatGPT Pro, Business Premium, and Enterprise users. Furthermore, only one access point can currently be configured, although OpenAI states that multiple access points will be available in the future.