That would be awesome. But it sounds like the llm has some access to customer accounts and call tools. Prompt injection remains a fundamentally unsolved problem, so I’m curious about how Intel plans to handle that.