AI Chatbots Could Enable Guidance on Biological Attacks, But More Testing Needed on Threats

AI chatbots could provide guidance to help plan and execute biological attacks, according to a new report.

The report by Rand Corporation tested large language models (LLMs) and found they could offer information to bridge knowledge gaps about biological agents.

LLMs discussed obtaining and transporting agents like those causing plague, as well as delivery mechanisms for toxins like botulinum.

The models recommended cover stories for acquiring dangerous bacteria while appearing to do legitimate research.

The report says more testing is needed, but LLMs may represent a new threat beyond what's already available online.