NPR reports on the privacy risks posed by OpenAI’s new web browser, which includes built-in access to ChatGPT and an agentic version capable of completing tasks for you. The company said it does not automatically use people’s information inputted into Atlas, but privacy advocates allege the browser’s ability to access your email or other online accounts puts data at risk. The AI agent is also threatened by prompt injections, where a website’s code can be manipulated to give the system instructions without the user knowing.
Full story