OpenAI launched its ChatGPT Agent on Thursday, its latest effort to turn AI into a profitable enterprise, as opposed to just consuming investors’ billions. In their announcement blog, OpenAI states that its Agent “can now do work for you using its own computer,” although CEO Sam Altman cautions that the rollout carries unpredictable risks.

AI agents are machine-learning tools aimed at executing multi-step tasks, representing a significant milestone in the AI arms race among competitors like Google and Microsoft. Demos ahead of the release showcased ChatGPT Agent fulfilling automated calendar management and generating financial presentations.

OpenAI reports that the Agent, a combination of earlier models, can carry out “complex tasks from start to finish,” typically completing tasks in 10 to 15 minutes, though more challenging tasks require longer.

Lisa Fulford, OpenAI’s research lead, mentioned using the Agent to order “a lot of cupcakes,” which took about an hour due to the specificity required. She remarked that it was more convenient than doing it herself.

However, Altman took to X to caution against using the Agent for high-stakes situations, stating: “I would explain this to my own family as cutting edge and experimental; a chance to try the future, but not something I’d yet use for high-stakes uses or with a lot of personal information until we have a chance to study and improve it in the wild.” He further advised against giving the Agent unfettered access to personal information or allowing it to manage sensitive tasks due to potential security vulnerabilities.

For those fascinated, the ChatGPT Agent is currently available for Pro users, with Plus and Team access rolling out shortly.