Rand兰德：2025年大语言模型双重用途生物能力自动化分级（AG）评估报告-高效技术框架与产业风险洞察（英文版）

Large language models are tools that can provide users with knowledge from diverse domains. Their ability to draw on the vast body of online knowledge and synthesize that information has led to concerns about the ability of large language models (LLMs) to provide dangerous information, such as instructions for building a biological weapon, a potential consequence of the dual-use nature of advancements in artificial intelligence (AI) and biotechnology. Prior evaluations have examined whether L