In a revolutionary move, OpenAI has unveiled the “ChatGPT agent,” a new feature that gives the chatbot the ability to use its own virtual computer to browse the web, download files, and interact with your applications independently. What does this mean for the future of artificial intelligence?
Introduction: a leap from assistance to autonomy
Don’t let OpenAI’s delayed open-source model slow it down in shipping other features. Today, the company is unveiling the “ChatGPT agent,” a feature that allows its chatbot to independently browse the web, conduct extensive research, and download and create new files for its human users using its own virtual computer. This means ChatGPT is no longer just a chat interface, but an agent capable of executing complex tasks in the digital world.
What does it mean for ChatGPT to have its own computer?
It might sound strange, but the idea is simple. Instead of being restricted to the chat interface, the agent can now use a set of tools that mimic what a human can do on a computer. OpenAI explains: “The model can choose to open a page using a text or visual browser, download a file from the web, process it by running a command in the terminal, and then render the output back in the visual browser. The model adapts its approach to execute tasks quickly, accurately, and efficiently.”
This gives the agent the ability to log into your accounts (like Gmail), download or send items on your behalf, all in a virtual and secure manner.
How to use it: from command to completion
Users can run the agent by clicking the “Tools” button in the ChatGPT command input box and selecting “Agent Mode.” Next, the user can describe a task in plain language, and the agent can execute it across web environments and local applications, combining reasoning and actions that only a human user could manually perform on their device.
The “ChatGPT agent” can connect to applications like personal or work Gmail and GitHub, so it can pull useful information—emails or code—from your accounts to assist with the tasks you ask it to perform. It can also connect to third-party APIs to pull information and use connected applications and services through them as well.
Merging Operator and Deep Research
The new “ChatGPT agent” builds upon and expands from the “Operator” agent released by OpenAI in January 2025, which allowed ChatGPT to browse the web. However, Operator was limited to interacting with websites only. In fact, the “ChatGPT agent” combines the best of “Operator” and “Deep Research,” another agent introduced in February 2025, which comprehensively searches the web to compile information into in-depth reports. The previous Operator tool will be phased out, but users can still access Deep Research.
Safety and control at the forefront
Because the agent can now take actions on behalf of users, OpenAI has introduced extensive safety measures. These include user confirmations before taking action, active supervision of sensitive tasks, and technical safeguards to limit unintended behavior. Key protections include:
- Confirmation prompts before actions like submitting forms or emails.
- “Monitor mode,” which pauses execution when the user becomes inactive.
- Refusal of high-risk tasks, such as financial transfers or privacy violations.
Conclusion: the future of work with artificial intelligence
The launch of the “ChatGPT agent” signals a shift in how users interact with artificial intelligence—from asking questions to assigning entire tasks. With its ability to reason, act, and produce outputs, OpenAI is betting that users increasingly want AI to work for them, not just help them. While the company emphasizes that the agent is still evolving, it sees this launch as the foundation for a more interactive and action-oriented future for AI.
