The rapid rise of frontier AI models is reshaping the debate over regulation in Washington as the U.S. and China compete for superiority [1].
This competition matters because the ability to develop and control the most powerful AI systems is now viewed as a critical geopolitical advantage. Policymakers fear that restricting access to these models will not stop bad actors from utilizing the technology, particularly as China advances its own capabilities [1, 3].
Giulia Neaher, a research analyst at The Stimson Center, said that the shift in model availability is changing how the U.S. approaches AI security [1, 2]. The debate centers on whether the government can effectively limit access to frontier models without hindering domestic innovation, or ignoring the reality of global open-source development [1, 2].
Recent data suggests the gap between proprietary and open-source systems is narrowing. Open-weight AI cyber capabilities now trail closed-model frontier systems by four to seven months [4]. This is a significant decrease from the previous gap of six to 10 months [4].
As these open-weight models close the capability gap, the effectiveness of U.S. export controls and access restrictions is being questioned [1, 5]. The White House is currently shaping policy on how to manage this access to prevent the technology from falling into the hands of adversaries [3].
Experts said that the proliferation of open-source models makes it nearly impossible to maintain a permanent lead through secrecy alone [1, 2]. The race has evolved into a geopolitical struggle involving chip theft and strategic shutdowns [3].
“Restricting access will not stop bad actors, especially as China’s open‑source models advance.”
The shrinking window between closed-source frontier models and open-weight alternatives suggests that technical superiority is becoming transient. For U.S. policymakers, this means that traditional 'moats'—such as restricted API access or hardware export bans—may provide only short-term advantages, shifting the strategic focus toward resilience and the ability to iterate faster than competitors.
