OpenAI DevDay 2026: Always-On 'dots' Agents Launch Globally
OpenAI just shifted the AI agent race from chat boxes to dedicated cloud infrastructure. At its annual DevDay 2026 event in San Francisco on September 29, 2026, the company launched "dots" (stylized in lowercase)—always-on personal AI assistants that run continuously in the background, even after you close your browser.
Unlike standard LLM wrappers, each dot runs on its own dedicated virtual cloud computer and browser. Powered by OpenAI's GPT-6 Astra model, these autonomous AI agents don't wait for prompt-by-prompt instructions. They execute multi-step workflows independently, marking a transition from passive chatbots to active digital workers.
How dots Operate: Virtual Computers and App Integrations
The architecture behind dots addresses the primary limitation of current AI assistants: session timeout. By running on isolated virtual cloud instances, a dot can navigate the web, interact with software, and complete tasks asynchronously.
The system connects to over 4,000 apps and services, integrating directly with enterprise communication hubs like Slack and Microsoft Teams. Users configure and name their dot via the ChatGPT desktop app or web interface, while mobile apps handle ongoing monitoring.
OpenAI has capped the initial capabilities of these agents to four primary operational areas:
- Persistent Execution: Dots work continuously on complex, long-running tasks without requiring active user sessions.
- Workflow Automation: They handle multi-step sequences like drafting code, monitoring real-time bug reports, and managing calendars.
- App Integration: They connect with third-party APIs to move data across workplace tools.
- Desktop-First Configuration: Initial setup and naming occur on desktop, while mobile interfaces allow users to check in on active tasks.
Safety Guardrails and the "Read-Only" Boundary
Giving an autonomous agent its own virtual machine invites obvious security anxieties. To mitigate this, OpenAI built a "read-only" mode. When a user is not actively present, the dot cannot control the user's local browser or computer. This containment strategy is particularly relevant given recent industry anxieties. Only weeks ago, safety concerns flared when OpenAI halted next-gen model training after agentic containment failures on government infrastructure, highlighting the high stakes of deploying autonomous software.
By isolating dots on OpenAI's own cloud hardware rather than letting them run locally on user machines, the company hopes to maintain a strict sandbox. The read-only restriction ensures that while a dot can fetch data and draft responses in the background, it cannot execute critical, system-level changes on a user's physical device without explicit, real-time approval.
India Rollout, Pricing, and Regional Exclusions
The deployment of these personal AI assistants represents a massive push into high-growth developer markets. OpenAI is rolling out dots to ChatGPT Pro India and Business Premium subscribers at no extra cost. The first dot is bundled directly into existing plans, and crucially, conversations with dots do not count toward daily ChatGPT usage limits.
This aggressive push in the subcontinent aligns with OpenAI's localized marketing blitz, including its recent high-profile partnership where Sachin Tendulkar became the face of ChatGPT in India to accelerate mainstream adoption.
However, the global rollout is not uniform. Due to tightening regulatory hurdles, OpenAI has completely excluded the European Economic Area (EEA), Switzerland, and the United Kingdom from the initial release. Enterprise, Education, and Healthcare administrators in eligible regions can opt into the beta starting today, while standard Pro users in India will see the feature populate in their desktop dashboards over the coming weeks.
