Malicious AI models could hijack the servers that run them

Original: LLMs could control their host machines by exploiting inference engines

Source: boydkane.com ↗