Artificial intelligence isn’t just a technology; it’s a geopolitical asset. The nation that leads in AI innovation will set the pace for economic growth, healthcare innovation and military capability.
The foundation of innovation has always been research. This is why senior defense officials and policymakers are pushing aggressively for more funding around AI R&D. But amid this flurry of policy and investment, protecting the AI discoveries that are so critical to national security advantage has gone overlooked.
Research security is the ability to safeguard the scientific enterprise itself — ensuring that the people, partnerships and data behind American innovation advance our strategic interests rather than those of our adversaries. In the race with China, this blind spot could prove decisive.
China takes the lead in AI research
Until very recently, the conventional wisdom was that the U.S. held a decade-long lead over the rest of the world in terms of AI innovation. China’s backing of the release of DeepSeek last year directly challenged that notion, and now the underlying data on global research trends paints an even starker picture. According to a new report produced by Digital Science CEO Daniel Hook, China now leads the world in AI research output, surpassing both the U.S. and Europe:
- China files nearly 10x more AI patents than the U.S.
- China’s output in AI research has eclipsed the combined total of the U.S., EU-27 and the UK in both volume and growth rate.
- China’s AI researcher base is larger and younger than what we have in the U.S. suggesting a continued wave of innovation.
- China has become the top AI research collaborator for the U.S., despite geopolitical tensions.
The fact of the matter is that China has become the top global collaborator in AI research, connecting its institutions into partnerships across continents — including with American universities.
On the surface, these collaborations appear to be the free exchange of scientific knowledge and discovery. In reality, they can serve as channels for the transfer of intellectual property, foreign influence or the quiet embedding of strategically distorted data into global research streams.
Consequently, the U.S. is increasingly dependent on a research ecosystem shaped and often led by its main global adversary. Outdated oversight systems make it difficult to grasp the extent of this reliance, creating a strategic gap where agencies must weigh the benefits of collaboration against growing national security risks.
The administration understands the stakes
In addition to the sweeping AI Action Plan, two recent administration directives underscore the urgency around research security.
Executive Order 14303 explicitly directs agencies to align science with U.S. national interest and security. This requires a level of control and vetting over research partnerships, data flows, and outcomes.
The OSTP’s Gold Standard Science initiative further requires agencies to ensure that research is transparent, reproducible, and free from bias, demanding deep visibility into the entire research lifecycle.
However, agencies are being asked to achieve this alignment while relying on fragmented and outdated tools. They struggle to answer even the most basic but vital questions:
- Who exactly are we funding?
- What affiliations — declared or hidden — do those researchers hold?
- Are there any indirect ties with foreign talent programs or adversarial institutions?
- How much of our AI progress depends on open-source models developed abroad?
Without answers, leaders face a false choice: embrace collaboration and risk exploitation, or shut down partnerships and risk isolation. Neither path is viable if America wants to maintain leadership in AI.
Why research intelligence is critical infrastructure
The solution isn’t more bureaucracy — it’s greater visibility. Agencies require integrated, real-time intelligence to map collaboration networks, verify expertise, and track global data flows as rigorously as we manage supply chains and defense systems. Research security must become a core pillar of national defense. Just as we would never deploy military hardware without a verified supply chain, we cannot afford to advance AI without full transparency into the sources of research, data and expertise.
Achieving this shift requires moving from reactive oversight to proactive vigilance. Agencies must:
- Map research networks to uncover hidden affiliations and dependencies
- Trace the provenance of data and models to protect against compromised sources
- Anticipate risks in emerging technologies before they become strategic threats
The tools to illuminate these connections already exist. What remains is for agencies, funders and policymakers to collectively recognize that research intelligence is now as vital to national security as cybersecurity or counterintelligence.
A defining strategic crossroads
The stakes for inaction have never been higher. Without strong research security, the U.S. risks unintentionally supporting adversarial agendas, embedding vulnerabilities in vital AI systems and undermining public trust in the integrity of science.
By treating research security as strategic infrastructure, America can collaborate globally with clear-eyed vigilance, accelerate innovation without blind spots and align the nation’s scientific enterprise with its broader security objectives.
Jürgen Wastl is the vice president of Research Evaluation and Global Challenges, Digital Science
