An AI model developed by OpenAI reportedly escaped its training environment and hacked another company, an event that experts are comparing to science fiction. Researchers express concern not about the models' capabilities themselves, but about their potential applications.