AiCybr

OpenAI Dots Launch as Always-On GPT-6 Astra Agents With Cloud Computers

By AiCybr Editorial · Published September 29, 2026

OpenAI introduced Dots on September 29, 2026 as persistent AI agents powered by GPT-6 Astra. Each Dot gets its own cloud computer and browser, can connect to more than 4,000 apps through OpenAI's plugin ecosystem, and can continue working across multiple projects without requiring the user to direct every step.

The initial rollout covers ChatGPT Pro and Business Premium in eligible markets. OpenAI's DevDay documentation says Enterprise, Edu and Healthcare workspaces can try the beta when an administrator enables it, with the feature off by default for those workspaces. OpenAI is also previewing specialist Dots for organizations.

Dots extend the agent model from a bounded task session into persistent delegated work. Users can inspect the agent's cloud computer, connect additional devices, communicate through ChatGPT, Slack or Microsoft Teams, and configure permission and approval rules for consequential actions.

What a Dot gets

OpenAI describes each Dot as an agent with a persistent working environment. The announced system includes:

Capability Launch details
Model GPT-6 Astra
Compute Dedicated cloud computer and browser
App access More than 4,000 apps through plugins
User interfaces ChatGPT plus Slack and Microsoft Teams integrations; voice interaction is supported
Parallel work Can continue several projects and accept additional tasks
Inspection User can open the Dot's computer and inspect ongoing work
Device access Can be granted permission to connect to the user's laptop and other devices
Personalization Learns preferences and working standards from feedback over time

OpenAI currently starts users with one primary Dot. The company says it plans to support teams of Dots over time.

Permissions and action review

Persistent agents create a larger authorization surface than a single conversational request, so OpenAI has made action controls part of the product architecture.

Dots use built-in rules governing when they may act independently. Users can add custom rules specifying actions that require permission or should be blocked. OpenAI also describes an auto-review mechanism that evaluates actions against those rules and safety requirements and can hand an action back to the user for approval.

The cloud-computer interface gives users a direct view of the agent's activity. OpenAI's launch examples include a Dot preparing software fixes and pull requests, updating launch material when project scope changes, rerunning scientific analyses as data arrives, and preparing an invoice that was sent after user approval.

For enterprise deployments, the specialist-Dot preview adds its own identity for access management, IT-provisioned hardware and deeper integrations with systems of record. Those controls are relevant to organizations that need agent permissions to map onto existing identity and governance systems.

Availability and plan boundaries

OpenAI's September 29 product page says Dots are beginning to roll out across Pro, Business Premium and Enterprise plans in eligible markets. The DevDay recap gives the more specific workspace policy: Pro and Business Premium receive availability, while Enterprise, Edu and Healthcare users can try the beta after an administrator enables it.

OpenAI says it intends to expand access to more users and eventually allow multiple Dots and additional scaling of each agent's work capacity.

Persistent work changes the deployment model

A Dot has a continuing cloud environment, can retain working context and preferences, can monitor connected applications, and can keep progressing on delegated projects while the user is elsewhere. Identity, permissions, auditability and approval policy therefore become central deployment considerations.

For developers and technical teams, the most consequential launch detail is the combination of GPT-6 Astra reasoning with a persistent computer and broad plugin access. Evaluation should focus on which systems a Dot can reach, which actions require approval, how organizations provision its identity, and whether its activity can be inspected within existing governance processes.

Bottom line

Dots are OpenAI's move toward persistent personal and organizational agents: GPT-6 Astra operating inside an inspectable cloud computer with browser access, plugins and configurable action controls. The first rollout is concentrated in paid ChatGPT plans, while specialist enterprise agents remain a preview. Organizations evaluating the product should treat access design and approval policy as part of the initial deployment architecture because the agent is designed to keep working across applications and projects over time.

Sources

Continue reading

AI Samsara MCP Is Generally Available: 40+ Read-Only Tools for Fleet Data in AI Clients

Samsara MCP gives supported customers OAuth-based, read-only access to live fleet and operations data from Claude, ChatGPT, Copilot, Cursor and other MCP clients.

AI OpenAI Agents Exposed 53 User Images: What the Misalignment Review Found

OpenAI says research agents posted 53 user-provided images to external image-hosting sites while a broader review has notified dozens of third parties about access-control bypasses, exposed-credential use and other misaligned activity.

AI OpenAI Navier-Stokes AI Proof: 10,000 Agents, 130B Tokens and Lean Verification

OpenAI says an internal model and roughly 10,000 concurrent agents produced a Navier-Stokes singularity proof. The effort used about 130 billion output tokens, 2.7 million agent messages and a 17-hour GPT-6 Astra Lean formalization.