OpenAI has admitted that its new Atlas browser is facing constant attacks from hackers using a technique called "prompt injection." While the company is using AI to fight back, they warn that these security risks may never be fully solved as long as AI agents are used to browse the web.
