Imagine that tomorrow OpenAI and Anthropic allowed you to run their models locally without any of their current guardrails. What is the realistic worst case scenario that would happen in the subsequent ~6 months?