US Government Control of AI Projects: 72% Probability
As the geopolitical struggle for artificial intelligence supremacy intensifies between the United States and China, the legal and political frameworks for direct government intervention are rapidly maturing. There is a 72% probability that the US government will exercise exclusive operational control over at least one non-military general-purpose AI project or its underlying model before 2029. This shift suggests a move from mere regulatory oversight toward 'nationalized AI dominance,' driven by the strategic necessity of protecting frontier model weights from foreign adversaries.
The Geopolitical and Strategic Driver
The urgent imperative for control is fueled by the perceived threat environment, specifically the US-China strategic rivalry. Evidence of industrial-scale theft and 'distillation' campaigns by Chinese actors has turned AI model weights into sensitive national security assets. Because China employs a 'military-civil fusion' policy, the US is moving toward a posture that treats frontier AI models as critical national infrastructure rather than private intellectual property. The logic suggests that if a private company cannot protect its models from being replicated by an adversary, the government may conclude that private ownership constitutes a liability.
A Robust Legal Arsenal
The US government possesses a formidable toolkit to facilitate this transition without necessarily writing new legislation. The Defense Production Act (DPA) allows the President to prioritize production for national defense and potentially secure government stakes in private firms. Additionally, the International Emergency Economic Powers Act (IEEPA) provides authority to block transactions or confiscate property during declared emergencies. Furthermore, by designating AI labs as 'supply chain risks,' the government can use CFIUS-style oversight to force divestitures or governance changes, effectively transferring operational control to government-approved entities.
Shifting Policy Landscapes
Recent political trends indicate a transition from establishing safety standards to prioritizing dominance. While earlier efforts focused on 'responsible AI' through frameworks like Executive Order 14110, newer proposals—such as the Trump America AI Act—aim to mandate that companies hand over code and weights to the Department of Energy as a precondition for deployment. This represents a 'de facto' transfer of control where the government secures the intelligence within the model even if it does not own the physical corporate entity.
Obstacles and Counterarguments
Despite the high probability, significant hurdles remain. The Fifth Amendment requires 'just compensation' for seized property, which could lead to massive litigation costs. There is also a risk of an innovation drain, where developers might move research offshore to avoid state seizure. However, the government may bypass these issues by pursuing 'operational control' rather than full ownership—using mitigation agreements or special security arrangements to manage specific high-value projects rather than entire corporations.
Conclusion: The Path Toward 2029
The groundwork for this shift is already being laid through legal readiness, strategic necessity, and a hardening bipartisan consensus on AI dominance. While total nationalization of trillion-dollar corporations faces constitutional barriers, the seizure of specific, high-value model deployment rights or weights is highly likely. By 2029, it is plausible that advanced intelligence will be managed by government agencies to ensure absolute American dominance in the field.
