AI 3 weeks ago OpenAI says AI browsers may always be vulnerable to prompt injection attacks OpenAI says prompt injections will always be a risk for AI browsers with agentic capabilities, like Atlas. But the firm…