OpenAI Launches Dots: Always-On Personal AI Agents
OpenAI has introduced Dots, a new kind of AI agent designed to stay active around the clock. The announcement came at the company’s DevDay 2026 event, and Dots are powered by OpenAI’s latest model, GPT-6 Astra.
How Dots Are Different From ChatGPT or Codex
Most AI tools, including ChatGPT and Codex, wait for you to give them a prompt before they act. Dots work differently. They are built to keep working toward a goal you set, continuing over time instead of pausing after each step.
OpenAI says each Dot gets its own cloud computer and browser. This means a Dot can keep working even after you log off, something that sets it apart from a typical chatbot session that ends the moment you close the tab.
What Can Dots Actually Do?
Users can create a primary Dot, give it a name, and assign it ongoing work. From there, Dots can:
- Work across connected apps
- Communicate through ChatGPT, Slack, and Microsoft Teams
- Eventually support text-message communication, according to OpenAI
OpenAI also wants to eventually let teams of Dots work together on a user’s behalf, rather than relying on just one agent for everything.

Real-World Examples OpenAI Gave
To explain how this might work in practice, OpenAI shared a couple of examples:
- A software developer could use a Dot to monitor customer feedback and start preparing fixes based on what it finds
- A scientist could use a Dot to rerun analyses and investigate unexpected results as new data comes in
Both examples point to the same idea: Dots are meant to handle ongoing, ambient monitoring-type work, the kind of task that would normally require someone to keep checking in manually.
Built for Businesses Too
Beyond individual use, OpenAI says businesses will be able to deploy specialist Dots that come with their own identities, credentials, and tools. This suggests a Dot could be set up almost like a limited digital employee, with defined access and permissions.
OpenAI is also working with Microsoft to integrate these agents with Agent 365 security controls, which points to a focus on enterprise-grade oversight rather than just consumer convenience.
The Model Behind Dots: GPT-6 Astra
Dots run on GPT-6 Astra, which OpenAI describes as its most capable model for complex reasoning, coding, research, and computer use.
Some key specs:
- 1.05 million-token context window
- Up to 128,000 output tokens
Benchmark Performance
OpenAI shared a set of benchmark results for GPT-6 Astra:
| Benchmark | Score |
|---|---|
| ARC-AGI-3 | 99.9% |
| FrontierMath Tier 4 | 97.6% |
| GPQA Diamond | 96.0% |
| BrowseComp | 91.5% |
| SRE-Bench | 88.0% |
| Terminal-Bench 4.0 | 57.9% |
| Agents’ Last Exam | 59.3% |
| AutomationBench | 41.4% |
Two safety-related figures stand out. OpenAI reports a 0% rate on its internal circumvention benchmark, compared with 0.29% for its earlier GPT-5.6 Sol model. It also reports a 1.8% rate on its computer-use safety benchmark with AutoReview enabled.
These numbers matter because an agent that operates autonomously, without a human checking every step, needs to be tested specifically on whether it tries to bypass restrictions or misuse its access to a computer.
Who Can Access Dots Right Now?
Availability is currently limited. Dots are available to:
- ChatGPT Pro and Business Premium users in eligible markets
- Enterprise, Edu, and Healthcare organizations, but only through a beta that an administrator needs to enable
How Much Control Do Users Have?
OpenAI says users can set rules that control what a Dot is allowed to do automatically versus what requires approval first. Sensitive actions, according to OpenAI, remain subject to explicit user consent, meaning a Dot shouldn’t be able to take major or risky actions without you signing off first.
Why This Matters
The shift here is fairly significant. Most AI tools today are reactive, they respond when you ask something. Dots are positioned as proactive, continuing to work on a defined goal without needing constant prompting. Combined with the safety benchmarks OpenAI is highlighting, this launch signals a push toward AI agents that can be trusted to operate with more independence, at least within limits the user defines.
Whether Dots deliver on this in practice, especially for everyday users rather than developers or researchers, will likely become clearer as access expands beyond the current Pro and Business Premium tiers.
