OpenAI has introduced a new AI agent for the ChatGPT platform, expanding the chatbot’s capabilities to manage more complex and multi-step tasks. The new agentic system builds on the company’s earlier features including Operator, which enabled ChatGPT to interact with websites, and its in house Deep Research offering, designed for carrying out in-depth, multi-step tasks.
With the update, these capabilities have been integrated into a single, more powerful agentic system dubbed ChatGPT agent, enabling the model to independently complete complex tasks by switching between reasoning and action.
The update allows ChatGPT to autonomously handle tasks such as preparing slide decks, summarising emails, analysing competitors, or booking travel. The agent can browse websites, run code, interact with APIs and generate editable content outputs like spreadsheets or presentations.
To deliver these services, the AI agent uses a virtual computer with web access and a suite of tools. It can also connect with third-party apps including Gmail and GitHub, enabling it to retrieve relevant information based on user prompts.
In a statement, OpenAI emphasised safeguards with the latest update aimed at upholding security, including ‘watch mode’ for active user supervision of critical tasks like sending emails, and ChatGPT agent seeking explicit user permission before executing consequential actions.
The upgrade is now available for Pro, Plus and Team subscribers, with enterprise and education users set to receive access in the coming weeks. The chatbot’s Operator research tool will also be sunset in the coming weeks as its functionality has been integrated into ChatGPT agent.
The latest update comes as AI agents rapidly gain traction across the tech industry. Major players including Microsoft, Salesforce and Oracle are investing heavily in the technology to enhance productivity and streamline operations.
Source: Mobile World Live
Image Credit: OpenAI