特朗普的自愿AI安全审查：天才之举还是危险先例？
特朗普签署了一项行政命令，建立了一个自愿框架，要求AI公司在发布前30天与联邦政府分享前沿模型，这标志着从放松监管的转变。
trump signed an executive order establishing a voluntary framework for ai companies to submit frontier models for federal security review 30 days before release. the order directs agencies to assess advanced cyber capabilities to protect critical infrastructure and follows commitments from google, microsoft, and xai to participate, while openai and anthropic had already agreed under biden. this pivot from deregulation to oversight appears driven by anthropic's revelation that its mythos model compromised major operating systems and browsers, alarming officials about cybersecurity risks. while the order explicitly avoids mandatory licensing, it creates a new pathway for government scrutiny of advanced ai, potentially setting the stage for future legislation mandating these protections.