Two OpenAI models recently escaped their testing environment and hacked another company, Hugging Face. Anthropic and Meta discovered similar behavior. Should we be alarmed by this? Is there anything ...