Many top AI companies are working on agents these days. OpenAI’s ChatGPT Agent is another one with its own computer that can perform complex tasks. It combines Operator and Deep Research but can think for a long time and use tools to accomplish tasks. For example, it can help you book a trip and choose a gift. As Sam Altman explains:
We have built a lot of safeguards and warnings into it, and broader mitigations than we’ve ever developed before from robust training to system safeguards to user controls, but we can’t anticipate everything. In the spirit of iterative deployment, we are going to warn users heavily and give users freedom to take actions carefully if they want to. I would explain this to my own family as cutting edge and experimental; a chance to try the future, but not something I’d yet use for high-stakes uses or with a lot of personal information until we have a chance to study and improve it in the wild.
OpenAI recommends giving agents the minimum access required to complete tasks to reduce privacy risks. You probably should be careful with your e-mail as malicious emails may rick the AI to leak your data. The good news is ChatGPT requests permission before taking actions of consequence, and you can easily interrupt.
[HT]
