- OpenAI launched ChatGPT Agent, allowing users to request detailed tasks like planning a wedding or online shopping—and the agent executes them independently.
- ChatGPT Agent uses a visual browser, terminal, and direct API access and integrates with services like Gmail and GitHub to efficiently perform tasks.
-
The feature is available only to Pro, Plus, and Team subscribers, with safeguards that prevent high-risk actions (e.g., bank transfers) to ensure user data safety.
OpenAI has unveiled a powerful new feature on its ChatGPT platform called ChatGPT Agent, designed to handle complex, end-to-end tasks—ranging from planning weddings to managing online shopping. The update marks a significant leap in the generative AI race and showcases OpenAI’s push to dominate the space.
The San Francisco-based company announced that ChatGPT Agent can process long, detailed queries and deliver multi-step results, allowing users to delegate time-consuming tasks more efficiently than ever before.
A New Era of AI-Powered Task Automation
In a detailed demonstration, OpenAI CEO Sam Altman showcased how the agent could take on elaborate tasks. For instance, the team asked ChatGPT Agent to plan a wedding, including outfit suggestions for five guests, gift recommendations under $500, and hotel options via Booking.com.
Within seconds, the agent responded with follow-up questions to gather missing details (like the wedding date), before delivering a comprehensive response covering every aspect of the request.
How It Works
ChatGPT Agent operates using its own computing environment, equipped with:
-
A visual browser to interact with websites using a GUI
-
A text-based browser for lightweight web queries
-
A terminal for code execution
-
Direct API access to connect with external platforms
It also integrates with third-party services such as Gmail and GitHub, enabling it to retrieve user-specific data when needed. Additionally, users can enable “Takeover Mode” to log in manually and allow the agent to go deeper into specific sites for research and execution.
“By giving ChatGPT multiple tools and interfaces, it can choose the most efficient route to complete any given task,” OpenAI explained.
Who Can Use It?
The new agent is available only to paid subscribers, including:
-
Pro ($20/month)
-
Plus
-
Team
The rollout began with Pro users, with others gaining access in the following days. Enterprise and Education tiers are still pending rollout. Notably, the feature will not be available in the European Economic Area or Switzerland at this time.
Safety and Privacy Measures
Since ChatGPT Agent can take actions on behalf of users and access sensitive information, OpenAI has implemented strict safeguards. For instance:
-
High-risk tasks like bank transfers are automatically blocked
-
Personal actions like sending emails require explicit user oversight
“ChatGPT Agent’s extended capabilities come with increased risk, especially when handling personal or sensitive data,” the company noted. “We will continue refining the tool to balance usefulness with safety.”
العربية (Arabic) To read the article in Arabic, click here
