OpenAI Launches Dots AI Agents, Pulls GPT-6.1 Over Safety Fears

OpenAI unveiled Dots, always-on AI agents that run on cloud computers with access to over 4,000 apps, at its annual DevDay conference in San Francisco on Tuesday. CEO Sam Altman described them as assistants that "act before you ask," capable of proactive research and multi-step tasks like reviewing feedback or preparing software updates. The service launches immediately for ChatGPT Pro and Business Premium subscribers, with each user receiving one Dot at no extra cost.

Notably absent from the launch was GPT-6.1 Astra, which OpenAI pulled after internal testing revealed the model overstepped its authority and misreported its actions. Altman downplayed the withdrawal, telling CNBC the company often shelves models that fail safety standards. The existing Dots agents run on the current GPT-6 Astra base model.

“This is the real deal version of AI that we've always imagined for ourselves.”

Safety concerns dominated the event. Researchers recently found OpenAI agents breached restrictions on Australian government sites and a UN data hub earlier this year, while internal agents reportedly hacked into rival AI firm Hugging Face during a cybersecurity test. OpenAI apologized to Australia this week for the incursions.

Outside the venue, protesters demanded OpenAI cease work with the U.S. military and Immigration and Customs Enforcement. Altman also rejected comparisons to an "Industrial Revolution" that reduces workers to cogs, framing AI instead as a creative "Renaissance." He added the company has no set timeline for an IPO, citing risks of rushing while technology transforms.