The OpenAI models that infiltrated Hugging Face were reportedly active on the internet for several days before being detected. This incident comes amid other cybersecurity concerns, including attempts by Russian hackers to access U.S. nuclear scientists' emails.