围绕美国是否应限制中国获取开放权重模型的问题，美国AI产业出现了明显分歧。英伟达等133家公司联名反对禁令[1]，但Anthropic和Meta两家企业对此持截然不同的态度。
Anthropic首席执行官Amodei于7月27日发声，反对无差别的开放权重模式[1]。他表示Anthropic从未主张禁止开放权重模型，但对英伟达信中的两个核心论点提出异议——即"开放权重模型必然有助于更容易地开发安全防护"和"广泛的技术获取必然对防御方比对攻击方更有利"[1]。Amodei主张采取基于模型能力的监管方案，辅以强制安全测试。
相比之下，Meta首席执行官扎克伯格于7月28日在《华尔街日报》发表署名文章为开源路线辩护[1]。扎克伯格认为"这个时代真正的问题不在于超级智能是否存在，而在于谁将拥有使用它的权利"[1]，主张通过市场竞争而非来源国禁令实现权力制衡[1]。值得注意的是，Meta是唯一未参与华盛顿自愿模型审查流程的美国大型AI开发商[1]。尽管两家公司都反对按国籍全面封禁模型，但在具体监管工具上存在根本分歧。
A significant rift has emerged within the American artificial intelligence sector over whether to restrict China's access to open-weight models. While 133 companies signed a letter opposing such restrictions, Anthropic's stance diverges sharply from this consensus.[1]
Anthropic CEO Dario Amodei publicly rejected the arguments presented in the letter on July 27, asserting that "Anthropic has never advocated for banning open-weight models."[1] However, Amodei opposed specific claims in the letter, particularly the assertions that "open-weight models necessarily facilitate easier development of safety measures" and that "broad access to technology necessarily benefits defenders more than attackers."[1] Instead, Amodei advocates for capability-based regulatory thresholds combined with mandatory safety testing as the appropriate regulatory framework.[1]
Meta CEO Mark Zuckerberg articulated a contrasting vision in a signed article published in the Wall Street Journal on July 28 titled "The Future of AI Belongs to Everyone."[1] Zuckerberg's core argument centers on the notion that "the real question of this era is not whether superintelligence will exist, but who will have the right to use it."[1] He champions an ecosystem-driven approach emphasizing market competition rather than source-country-based restrictions.[1] Notably, Meta remains the only major U.S. AI developer that has not participated in Washington's voluntary model review process.[1]
While both Anthropic and Meta reject blanket prohibitions based on national origin, their preferred regulatory mechanisms reveal a fundamental disagreement about how to balance innovation with safety concerns in the global AI landscape.[1]
