OpenAI has acknowledged that prompt injection attacks remain a persistent security threat for AI-powered browsers, even as ...