U.S. AI policy has effectively abandoned any semblance of a centralized strategy, devolving into a patchwork of clashing interests. Today, the fate of tech exports and sanctions depends less on established doctrines and more on whose office is situated closest to the Oval Office. While Chinese open-weight models gain momentum, the Trump administration has delegated the response to a fragmented group of officials with diametrically opposed views. As a senior White House official told WIRED, the internal debate is not a clear-cut choice between protectionism and growth, but a chaotic "ten-sided argument." For businesses and investors, this marks the end of regulatory stability: systemic rules have been replaced by personal lobbying.
Commerce Levers and the Benchmark War
Commerce Secretary Howard Lutnick has solidified his role as a power broker, transforming the Bureau of Industry and Security (BIS) into his primary instrument of control. Lutnick is attempting a delicate balancing act: imposing export restrictions on U.S. firms like Anthropic to "ensure compliance" while simultaneously seeking ways to encourage American labs to release their own open-weight models to undercut Chinese competitors. According to a WIRED report, Lutnick has personally courted AI lab executives to push this counterweight strategy. Publicly, he maintains a stance of performative skepticism; in a post on X, the Secretary trashed Moonshot AI’s Kimi K3 model, claiming his team found nothing remarkable in it compared to American flagships.
Lutnick is utilizing the BIS as a leverage point to pressure private AI laboratories. The Commerce Department is balancing between restrictive bans and supporting open-source models. Public criticism of Chinese developments has become a formal pillar of state rhetoric.
Working under Lutnick is Arvind Raman, acting director of the Center for AI Safety and Innovation (CAISI). Raman took the post following the abrupt resignation of Chris Fall, who had spent weeks trying to strong-arm Anthropic into implementing rigid safety protocols for its Fable 5 model. This friction between CAISI and the private sector proves that federal agencies are no longer passive observers. Per WIRED data, CAISI specialists effectively "moved in" to Anthropic’s headquarters, scrutinizing safety architecture before Fable 5 was permitted to go back online. The Commerce Department isn't just issuing licenses anymore; it is looking under the hood of the models themselves.
Hawks and National Security
On the opposite end of the spectrum is National Cyber Director Sean Cairncross, who advocates for an aggressively hawkish policy toward Chinese labs. A former campaign lawyer with no technical background, Cairncross was given carte blanche under Trump’s June 2 executive order. His mandate is to mitigate national security risks linked to AI. Despite his lack of specialized education, subordinates praise his willingness to dive into the era's most complex security issues. However, this hardline stance constantly hits a wall of resistance from free-market-oriented interest groups.
The result is an environment where a single social media post or one closed-door meeting can flip the nation's export policy 180 degrees overnight.
The administration pays lip service to a united front and a goal of doubling the pace of innovation to widen the gap with China. In reality, the lack of coordination has birthed a system where technical standards are bartered in backrooms and export bans are decided by lawyers who cannot distinguish model parameters from weights. It is framed as a strategy for dominance, but in practice, it is a high-stakes game of lobbying where the rules change as quickly as the names on the President's schedule.