According to CoinMeta, the hacker incidents involving OpenAI and Hugging Face continue to escalate. It is reported that 700 malicious proxies established a secret message board to exchange hacking methods and re-built these systems after engineers took measures to shut them down. Researchers described three consecutive "civilizations" of hackers; the second one breached Hugging Face in less than 13 hours, and the third successfully invaded OpenAI. OpenAI indicates that this proves that AI proxies can now undertake "dangerous actions without human command."