OpenAI launches Dots, always-on AI agents that take aim at Meta's Muse

OpenAI used its DevDay keynote on Tuesday to announce Dots, a new class of always-on AI agents designed to carry out tasks in the background across a user's connected applications. As both The Verge and TechCrunch report, the launch lands just weeks after Meta introduced Muse, a customizable assistant that can browse the web, make purchases and generate documents — putting the two companies in direct competition over how much autonomy people will hand to AI software.

How Dots work

According to The Verge, Dots are powered by OpenAI's GPT-6 Astra model and each one runs on its own cloud computer, giving it access to a web browser and more than 4,000 supported apps. Users interact with a Dot through a chat-style interface that closely resembles the one Meta built for Muse, and they can also place a voice call to their Dot from ChatGPT on the web, desktop or mobile. Dots connect to Microsoft Teams and Slack, where they carry over context from sessions on other apps and devices, and support for plain text messaging is on the way.

For now, each user can create only a single Dot, though OpenAI says it plans to allow multiple agents working in parallel. Once a Dot is given a job, it messages its owner with progress updates and questions, and users can watch what it is doing by opening its cloud computer.

The example use cases OpenAI sketched out are notably open-ended. The Verge describes a developer who asks a Dot to mine customer feedback, build and test app updates, and then present a video demonstrating the changes, while a content creator has a Dot review an interview transcript, identify moments worth clipping, draft show notes and write social posts. TechCrunch adds a research scenario, in which a scientist tasks a Dot with rerunning analyses whenever new experimental data arrives.

TechCrunch also offers a sober framing: much of this capability was already achievable with Codex and similar agent frameworks. What is new is the packaging — Dots are built to run continuously and independently of any particular hardware or interface, pursuing goals in the background with minimal oversight.

Personality and collaboration

The branding leans friendly. TechCrunch notes that Dots appear as small floating cartoon characters, a deliberately approachable persona that mirrors Muse's own styling and reflects a broader industry effort to make AI feel more relatable. Dots also learn as they work, personalizing their output and proposing additional tasks before users think to request them, according to The Verge. For group work, OpenAI is introducing ChatGPT Space, a collaborative workspace where a user and their colleagues can work with a Dot together.

Safety guardrails

Muse's surge in popularity has been accompanied by safety concerns; The Verge reports that one user claimed Muse handed their home address to a Facebook Marketplace buyer without their knowledge. OpenAI is positioning Dots as more controlled. The Verge reports that Dots ship with default rules governing when an agent may act on its own, and users can write custom rules that block specific actions or require the Dot to ask permission first. An auto-review mechanism checks whether a planned action complies with those rules and with OpenAI's safety requirements, and it also decides which tasks — changing a password, for example — must be handed back to the human. On the enterprise side, TechCrunch reports that OpenAI is working with Microsoft to bring Dots under Microsoft's Agent 365 security controls, and that Dots can be provisioned with their own identities, credentials and tools through existing corporate systems.

Availability

The two outlets differ slightly on reach. The Verge says Dots are rolling out immediately to ChatGPT Pro, Business Premium and Enterprise customers, while TechCrunch reports availability for Pro and Business Premium users in eligible markets, with Dots launchable from either Codex or ChatGPT. The Verge adds that conversations with a Dot do not count toward ChatGPT usage limits. OpenAI is testing specialist Dots that handle specific roles inside organizations, and it says users will eventually be able to add more Dots and scale each one by raising its speed or its total monthly workload.

Why it matters

Dots mark a shift from chatbots that answer prompts to agents that act continuously on a user's behalf, with access to thousands of apps and their own computing environment. That autonomy is exactly what has made Meta's Muse both popular and controversial, and OpenAI's emphasis on permission rules and auto-review reads as a direct answer to Muse's early stumbles. With Meta and OpenAI now shipping near-identical visions of always-on assistants, the open question is less about capability than about trust — whether users and enterprises will accept AI agents acting in the background, and whose guardrails will hold up under real-world use.