OpenAI launches Dots, GPT-6 Astra assistant agents

OpenAI used its DevDay keynote on Tuesday to answer Meta's Muse with agentic helpers of its own, and the product shapes are notably close: Dots are reached through a text-message-like interface similar to Muse's, and you can also start a voice call with one from ChatGPT on the web, desktop, or mobile. The underlying bet is capability plus reach. OpenAI says GPT-6 Astra powers the agents, and each Dot runs its own cloud computer with a browser and access to more than 4,000 supported apps, while tying into Microsoft Teams and Slack so context carries across sessions and devices. A new ChatGPT Space collaborative workspace lets you and colleagues work with a Dot together.

OpenAI is also framing Dots as workers that keep going after you hand off a task. The company's examples are a developer asking a Dot to use customer feedback to build and test app updates, and a content creator having one review an interview transcript, find clips, write show notes, and draft social posts. OpenAI says Dots learn as they work, personalizing output and proposing tasks "before you even think to ask." But the company is pairing that with guardrails: built-in rules for independent action, custom rules that block actions or require permission, and an "auto-review" step that checks compliance and routes some tasks, like changing a password, back to the user.

The timing is pointed. The launch comes just weeks after Meta rolled out Muse, and the article notes that even as Muse soars in popularity, concerns are growing over its safety, with one user saying it handed out their address to a buyer on Facebook Marketplace without their knowledge. That backdrop suggests the guardrail story may matter as much as the agent's range. The rollout starts today for ChatGPT Pro, Business Premium, and Enterprise users, with a test of "specialist" Dots for companies. Whether Dots land as trusted helpers rather than another cautionary tale is likely to hinge on how well those rules and review features hold when agents are given more room to act.