OpenAI plans to release a new artificial intelligence tool, codenamed Operator, that can autonomously use a computer on a person’s behalf. According to a Bloomberg report, potential uses for OpenAI Operator include writing software code or booking travel. An internal staff meeting from OpenAI leadership announced that Operator is scheduled to be released as a public preview in January 2025.
Operator will be accessible to research developers and users with an application programmable interface (API). It’s unknown if the tool will be limited to an invitation-only basis or released in a wide public preview.
OpenAI declined to comment on the matter when contacted by Bloomberg.
The news story broke the same day OpenAI released an artificial intelligence infrastructure blueprint for the United States government.
The policy document outlines OpenAI’s vision for “American AI” and was shared with the outgoing Biden administration in November 2024.
It’s unclear what stance the incoming Trump administration will take on OpenAI’s blueprint or generative AI.
Agentic AI is the new frontier in artificial intelligence
The announcement comes as many artificial intelligence and tech companies are moving beyond chatbot-like generative AI tools into agentic AI. Agentic AI provides greater utility than generative AI conversational or content-generation chatbots, such as Gemini by Google or ChatGPT by OpenAI.
Using either Gemini or ChatGPT requires using natural language prompts to instruct the large language model with specific tasks. In contrast, agentic AI mimics the human-like behavior of computer use for completing repetitive tasks or processing what’s occurring on a computer screen in real time.
Agentic AI uses sophisticated reasoning and logical planning to complete and solve multi-step problems autonomously.
Over time, agentic AI models can create a data flywheel to enhance efficiency and refine the model and outcomes from continuous data interactions.
Agentic AI: using computers like humans do
In October 2024, Anthropic released Computer Use, a new tool similar to OpenAI’s Operator. According to Jared Kaplan, co-founder and chief science officer of Anthropic, it is the first model capable of using a computer like a human.
Rather than rely on multiple tool integrations, Anthropic’s Computer Use can understand what is displayed on a user’s computer screen. With a user’s permission, it can browse the web, click buttons, and type autonomously.
Other companies, such as Microsoft and Salesforce, are working on their own agentic AI solutions, each with unique capabilities that will integrate into their respective platforms.
It is the first step in blurring the lines between humans and AI for everyday computer use.
However, it also poses significant challenges for future computer forensics and cybersecurity, where it could be difficult to ascertain whether a human or AI agent performed a task.
Disclaimer: The author of this article is a current employee of Google. This article does not represent the views or opinions of his employer and is not meant to be an official statement for Google, Google Cloud, or the Alphabet holding company.
