OpenAI launched ChatGPT Agent on 17 July, enabling users to delegate complex, real-world tasks to AI, including restaurant reservations, grocery shopping, and presentation creation. ChatGPT Agent represents an advanced version of the previous Operator tool, combining Operator's website interaction capabilities, deep research's information synthesis skills, and ChatGPT's conversational intelligence into a unified system.
The service operates within a virtual browser environment, allowing AI to interact with websites, fill out forms, and execute multi-step tasks autonomously. OpenAI Pro users receive 400 messages monthly, whilst Plus and Team users get 40 messages per month, with additional usage available through credit-based options. However, the company has identified new security risks, particularly regarding prompt injection attacks, where malicious instructions embedded in webpages could trick the AI into sharing sensitive data or performing harmful actions. To mitigate this, OpenAI requires explicit user approval before consequential actions and active supervision for critical tasks such as sending emails.
ChatGPT Agent began rolling out to Pro users by end of day on 17 July, with Plus and Team users gaining access over the following days, and Enterprise and Education users receiving access in the coming weeks. The service can integrate with third-party services like Gmail and Google Drive, enabling seamless email and document management. OpenAI emphasised that whilst significant safety measures have been implemented, the service remains in early stages and may make mistakes, particularly in slideshow generation.
Sources:
1.
OpenAI · July 17, 2025
2.
Julia Williams
3.
Antonio Pequeño IV
