Hackers and other criminals can easily commandeer computers operating open-source large language models outside the guardrails and constraints of the major artificial-intelligence platforms, creating security risks and vulnerabilities, researchers said on Thursday. Hackers could target the computers running the LLMs …
