OpenAI’s ChatGPT has evolved from a basic chatbot to a sophisticated agent, available for paying subscribers, enabling it to perform complex tasks autonomously. Announced in a recent blog post, this enhancement allows users to command ChatGPT to execute tasks like “build a cash burn rate model” using its own computing resources. Subscribers of Pro, Plus, and Team plans currently have access, while Education and Enterprise users will gain features in the coming weeks.
The ChatGPT agent incorporates advanced tools, including web browsing capabilities, content access, and integration with services like Gmail and GitHub. However, this functionality introduces new risks, such as potential data breaches through malicious prompts. OpenAI emphasizes the importance of safety controls, claiming enhanced defenses against adversarial prompt injections, and safeguards against harmful actions. The company has pledged to require user permissions for high-stakes tasks, ensuring that sensitive operations, like money transfers, remain secure.
Source link