OpenAI has unveiled the ChatGPT agent, a cutting-edge AI tool designed to execute complex tasks directly within users’ browsers. Leveraging a new reasoning-optimized model, it surpasses previous AI systems in a variety of benchmarks. This versatile agent automates tasks across multiple cloud applications, such as downloading code from GitHub and saving it to Google Drive, while also running security checks like vulnerability scans.
The ChatGPT agent employs dual browser functionality: one optimized for text-processing and another that interacts with graphical user interfaces. It prioritizes user consent for sensitive actions, ensuring tasks can be paused or re-instructed as needed. In addition to standard web applications, it can manage terminal commands for file editing.
OpenAI has implemented extensive safeguards against misuse, particularly against prompt injection attacks. Currently, ChatGPT agent is accessible via the Pro, Plus, and Team tiers, demonstrating superior performance in tasks such as mathematical reasoning and spreadsheet management.
Source link
