OpenAI says prompt injections will always be a risk for AI browsers with agentic capabilities, like Atlas. But the firm is beefing up its cybersecurity with an “LLM-based automated attacker.”

LEAVE A REPLY

Please enter your comment!
Please enter your name here