Read Part I: Control vs. Innovation here.
Part II: Why the Acquisition System Worked This Way—and What Are the Possible Consequences of the Change?
To understand the significance of this moment, it is important to understand what is being broken.
For decades, federal acquisition policy has operated on a fundamental premise: the government does not need to own technology to benefit from it. Instead, it licenses commercial products while allowing vendors to retain their intellectual property and reuse it across markets.
This model emerged for practical reasons. By allowing companies to keep their IP, the government ensured that private firms had strong incentives to invest in research and development. Vendors could spread development costs across multiple customers, continuously improve their products, and bring those improvements back to government users. The result was a mutually reinforcing system in which the government gained access to increasingly advanced technology without bearing the full cost of creating it.
This approach was especially critical in software. The rise of commercial off-the-shelf products and later cloud computing depended on vendors maintaining ownership of their systems. Federal policy was designed to align with that reality, not override it.
The GSA AI clause marks a departure from that model. It does not eliminate vendor ownership of base systems, but it significantly expands government claims over outputs, improvements, and derivative value. In doing so, it shifts the balance of incentives.
Under the traditional model, vendors benefited from government use because it contributed to product improvement and broader market value. Under the proposed AI framework, that benefit is sharply limited. Government use becomes more isolated, less reusable, and therefore less economically attractive as a driver of innovation.
This shift creates clear winners and losers.
The likely winners are those positioned to operate within a sovereign, government-controlled environment. Large integrators, compliance-focused firms, and providers capable of delivering secure, isolated AI deployments stand to gain. These players can absorb the regulatory burden and build offerings tailored specifically to government requirements.
The likely losers are those whose business models depend on scale, reuse, and continuous learning across customers. Frontier AI companies, in particular, may find that the federal market no longer supports the same level of participation or investment. Smaller firms without the resources to manage compliance complexity may also struggle to compete.
The broader impact is not the disappearance of innovation, but its redirection. Instead of a market driven primarily by technological advancement, the federal AI space may become one shaped more heavily by compliance, governance, and control.
That is not inherently a failure. It reflects a conscious decision to prioritize sovereignty and mission assurance in a domain where the stakes are unusually high. But it does come at a cost.
The historical model worked because it allowed the government to harness the full momentum of private-sector innovation. The emerging model seeks to contain and direct that momentum. Whether it can do so without slowing it—and whether that trade-off is sustainable in a competitive global environment—remains an open question.
What is clear is that the rules are changing. The Anthropic dispute is not an isolated incident, but an early indicator of a broader realignment. The future of federal AI will not simply be negotiated through contracts. It will be defined by how the government and industry resolve this fundamental tension between control and innovation—and by how well they can bridge the gap between them.
