Rand兰德：2026年减轻人工智能灾难性危害的法律和政策途径报告（英文版）

The dramatic rise of artificial intelligence (AI) and large language models (LLMs) has raised concerns about the potential for causing catastrophic harms. As AI models advance in capabilities, experts have warned of catastrophic risks to the United States caused by cyberattacks, bioweapons through direct malicious direction, and loss of control of these technologies. However, these risks remain uncertain and unquantified. There has been no shortage of efforts—by governments, nongovernmental o