前 NIST 官员 Jake Taylor 呼吁美国 AI 政策补齐“验证不对称”：白宫今夏辩论 AI 政策之际，OpenAI、Anthropic 和 Meta 的系统均在测试中入侵第三方系统，另有一款 OpenAI 系统产出附机器可检验证明的新数学成果。他认为 AI 能力增长快于核查能力，主张 CAISI/NIST 效仿半导体在线计量学，公开一套前沿模型评测层级并内嵌形式化推理检查，由派驻实验室的工程师建设，首批覆盖生物防御、儿童安全和合成材料溯源等领域。
The White House issued Executive Order 14409, "Advancing Advanced Artificial Intelligence Innovation and Security," on June 2, directing federal agencies to harden systems against AI-enabled threats, assess advanced cyber capabilities, and establish a vulnerability information sharing center, while requiring the Department of Justice to make using AI to commit cybercrime an enforcement priority. This Tech Policy Press analysis argues the order has transparency and accountability gaps: the benchmark process for determining whether a model meets the "covered frontier model" threshold will be classified, with the NSA Director making the final determination, the assessment framework is voluntary, there is no requirement to submit an unclassified summary report to Congress, the Government Accountability Office is not involved, and the public cannot obtain benchmark standards or assessment results through the Freedom of Information Act. The article notes the order focuses only on cyber risks and does not address the large-scale scraping of personal information when training frontier models, the profiling and inferential capabilities these models create, and their privacy implications.
Trump delayed signing an executive order on AI and cybersecurity, saying it "could be an obstacle," while according to The Washington Post, Zuckerberg, Musk, and David Sacks each spoke with the president the previous night and into that morning. A draft leaked to POLITICO shows the order is voluntary in principle, explicitly prohibiting interpretation as establishing mandatory licensing or pre-review, and only requiring the Department of Justice to prosecute, under the Computer Fraud and Abuse Act, those who use AI to illegally access or damage computers. The draft also grants the NSA authority to conduct classified evaluations of frontier models, with ONCD, CISA, and NIST (including CAISI) assisting in defining the scope. The article argues this arrangement shifts evaluation leadership from civilian regulators to intelligence agencies, whose processes are not subject to FOIA and whose results cannot be made public, and CAISI is reportedly already forming an interagency working group to conduct classified evaluations.
