Memo to Trump: Shape AI’s future by embracing open-source development

MEMORANDUM FOR THE PRESIDENT

FROM: OZAN AHMET CETIN, SETA WASHINGTON D.C.

PURPOSE: MAINTAINING OPEN-SOURCE AI

Past US administrations, including the first Trump presidency, have recognized the strategic importance of artificial intelligence and allocated substantial resources to its advancement. The Pentagon alone has been working on more than 800 AI-related projects. Thanks to significant funding and a robust innovation ecosystem, American companies are at the forefront of the field.  Hence, the United States can reasonably be considered a leader in artificial intelligence. But the rapid pace of AI innovation means the half-life of breakthroughs and their market position is remarkably short, and the journey toward the technology’s maturity—if such a point is ever reached—remains long and uncertain. A critical factor shaping AI’s trajectory will be how leading states and companies approach open-source development. The Trump administration should consider maintaining the open-source-friendly attitude of the previous administration.

Open-source AI refers to artificial intelligence tools, including models and pre-trained parameters, made freely available for public use and modification. Such openness enables developers to directly fine-tune or extend AI models without replicating the resource-intensive training process. The tech industry has embraced the open-source model for decades—starting well before the rise of advanced AI technologies—as a foundation for collaboration, transparency, and shared progress. From early operating systems like Linux to widely used programming languages and frameworks, open-source initiatives demonstrated how collective contributions could lead to robust, scalable, and adaptable technological solutions.

Despite the positive track record of past open-source efforts, the allure of closed AI persists, driven by fears of misuse by malicious actors and the risk of competitors gaining access, potentially undermining strategic advantages. These fears and risks are exaggerated. First, from a safety perspective, open-source AI benefits from constant scrutiny by an expert community, identifying vulnerabilities, mitigating malicious uses, and offering fixes. This collaborative style arguably reduces bias and helps prune pathways to potential existential risks. Second, the concern over open-source AI enabling competitors like China to gain access to powerful models misses the broader picture because it underestimates the dynamic and evolving nature of AI as a technology. AI is far from a mature field; its rapid innovation cycles and changing applications mean that today’s advantages may quickly become obsolete. Attempting to protect current leads through exclusivity could be a shortsighted strategy that cedes the global AI narrative to others. The true competition in AI is not merely about technological supremacy or who builds the most advanced models, but in shaping the global AI ecosystem. The critical questions revolve around whose architecture, infrastructure, standards, ethical frameworks, and overarching vision will guide the development of AI and influence its implementation on a global scale.

Moreover, open-source AI paves the way for a more equitable distribution of the value generated by artificial intelligence, preventing its concentration among a limited number of proprietary technology owners. Open-source AI significantly lowers barriers to adoption by making advanced technologies accessible to a broader range of users, from small startups to large enterprises—and even individual developers. As more entities integrate AI into their operations, the cumulative impact contributes to greater economic growth and development. In essence, open-source AI acts as a rising tide that lifts all boats, enabling collective progress while cultivating a more inclusive and dynamic technological landscape.

Recognizing the importance of this issue, the National Telecommunications and Information Administration (NTIA), under the Biden administration, defended openness in AI following comprehensive consultations with stakeholders last year. But there is also momentum in Congress to introduce additional restrictions on AI due to fears that China might gain access to US advances. This approach falls short of capturing the full scope of the issue. As META’s founder and CEO, Mark Zuckerberg, a prominent figure in open-source AI, notes “AI development will continue to be very competitive, which means that open sourcing any given model isn’t giving away a massive advantage over the next best models at that point in time.” Hence, the United States must balance its strategic concerns with the broader benefits of openness. Restrictive measures risk isolating American advancements and ceding influence to actors willing to adopt openness as a means of shaping the AI ecosystem.

To maintain AI leadership while embracing openness, the United States should focus on international collaborations, secure funding for open-source projects, drive secure AI research and development, and encourage industry participation. Establishing a global consortium for AI standards with allies can shape ethical and technical benchmarks. Federal grants and secure licensing models can support open-source innovation while mitigating misuse. Investments in privacy-preserving AI and public-private infrastructure partnerships will sustain a competitive edge. Additionally, incentivizing private-sector contributions through tax credits and a corporate open AI council can foster collaboration, accelerate innovation, and align efforts with national priorities.

Bulletin of the Atomic Scientists