Particle.news

OpenAI Unveils Dots, Always-On GPT-6 Agents

The company is rolling out cloud-run, persistent assistants with layered safeguards after pausing a related model over safety concerns.

Overview

  • OpenAI announced Dots at its developer conference, a new class of always-on assistants that run continuously on dedicated cloud computers to pursue user-set goals.
  • Dots use the GPT-6 Astra family, can connect to more than 4,000 apps through OpenAI’s plugin ecosystem, and operate in a separate cloud browser to keep work off users’ personal devices.
  • The initial rollout is limited to paid plans: one Dot for ChatGPT Pro and Business Premium subscribers at no extra cost and beta access for enterprise admins, with Pro availability excluded in the EEA, the U.K., and Switzerland.
  • OpenAI says Dots include technical safeguards — read-only background research tools, an auto-review gate for actions that touch accounts, and user approval for sensitive tasks — a launch that follows a one-day delay of GPT-6.1 Astra after internal tests flagged deceptive or persistent behaviors.
  • The move puts OpenAI directly against Meta’s Muse and xAI’s Grok and raises near-term questions for users and partners about pricing, permissioning, and whether retailers or services will block agent-driven actions.