OpenAI’s new Dots agent failed to respond during a live demonstration at its DevDay conference, prompting the presenter to say, “I guess Dot’s having a slow morning.” OpenAI’s Thibault Sottiaux acknowledged the problems: “Live demos suffered from rolling out all the updates at the same time.” He said the company remained committed to live demonstrations.

The GPT-6 Astra agents run on their own cloud computers and work across connected apps between conversations. Early tester Dan Shipper of Every reported mixed results. His Dot flagged a conflict between a new flight time and an unread meeting request in Slack, but he also encountered reliability problems. “I had frequent issues with permissions and dropped messages, and my Dot was unable to connect to its own in-app browser,” he wrote.

The launch also brought scrutiny of the agents’ safeguards. In OpenAI’s tests, scope-violation flags rose from 8.6% to 19.7% as task chains lengthened, Vector News reported. Dots retain information learned from an app after it is disconnected. Background research uses tools that cannot change content, while actions affecting accounts or sharing information go through automated review, custom rules and built-in safety checks.