This content was produced by Axios Creative House in partnership with L3Harris Technologies and was not created by Axios newsroom staff.

By Heidi Wood

Vice President, Chief Digital and Transformation Officer, L3Harris Technologies

With artificial intelligence evolving at an exponential pace and permeating nearly every aspect of our economy and daily life, regulators and industry leaders are debating what, if any, guardrails should be placed on it — particularly foreign AI models that could pose security and economic risks to the United States.

The outlook: While the concerns driving these discussions are legitimate, the conversation needs to be reoriented.

• Rather than debating what should or shouldn't be banned, our nation should be focused on how we can ensure America has a competitive advantage in every aspect of AI — including in "open-weight" models that make AI widely accessible and customizable.

From my perspective in the defense tech industry, ceding leadership in open-weight AI models to foreign companies is not a viable option. As we've seen in recent conflicts such as the war in Ukraine, modern battlefield challenges demand the application of open-weight models — systems whose trained parameters can be downloaded, adapted and run on infrastructure the user controls.

• Last year, Reuters reported that Ukraine deliberately avoids reliance on proprietary systems so they can integrate AI into battlefield management systems for troop coordination and enemy monitoring.

The strategy: To deter and defeat increasingly capable adversaries, American and allied warfighters need technologies built around models we can trust.

• Today, the world's strongest open-weight models originate from our nation's chief adversary: the People's Republic of China.
• As of July 2026, Chinese models occupied 22 of the top 25 positions on the Artificial Analysis Intelligence Index of open source models — a widely cited independent benchmarking tool for AI.

To be clear, our nation is in a position of strength in the overall AI race. American industry builds the most capable models in the world, and our government's own evaluations confirm the frontier still runs through this country.

• The National Institute of Standards and Technology recently found that open-weight AI models from the Chinese developer DeepSeek lag behind America's most advanced general-purpose models — known as "frontier" models — in performance, cost, security and adoption.

The Trump administration also deserves credit for making American AI leadership a top national priority. Their AI Action Plan released in July 2025 calls for leading open models founded on American values.

• They've also launched the Gold Eagle initiative — a voluntary public-private clearinghouse for finding and fixing software vulnerabilities.

Leadership at the frontier, however, is only part of the equation. The gap in open-weight AI between the U.S. and China reflects a choice not yet made rather than a capability we lack.

There are several actions our nation should take immediately to position America for AI dominance.

1️⃣ First, invest in a trusted, secure American ecosystem that drives open-weight innovation and make it more financially accessible to America's industries.

• Federal matching funds should flow to the companies, universities and research teams that train capable American open models through recurring competitions with clear criteria for capability and security.
• The competitions should be open to models of different sizes — not only the largest systems but also the fast, efficient ones built for the edge.

• The government should make advance commitments to license models that win the competitions, and participants should get to keep commercial rights to what they build on top.

2️⃣ Second, give open-weight AI its own "Blue List" — a U.S.-administered, voluntary model assurance registry where approval moves quickly and bureaucracy stays light, with a public catalog for broad use and controlled tiers for national-security missions.

• The instruments already exist across government. NIST, for example, holds the country's best measurement science for model security.
• Connecting them into one public repository would allow models to be certified for provenance, training-data lineage, resistance to hijacking and jailbreaking, and performance in disconnected environments, with signed artifacts and periodic re-evaluation.

3️⃣ Third, "Blue Listed" models should receive preference and a faster path to authorization across federal and defense programs, with agencies recognizing common test results instead of repeating the same review.

• Once the certified list becomes the doorway to the government market, developers will compete to be on it, which will drive greater innovation and security.

The takeaway: Leaders in Washington and Silicon Valley are right to be concerned about America's position in the global open-weight AI race. It poses a threat to not only our economy but also our national security.

• To maintain our technological advantage over our adversaries, America's warfighters must have access to capabilities built on trustworthy open-weight models.

Fortunately, our country retains every meaningful advantage in this contest — the best researchers, the deepest capital markets and the strongest AI labs in the world. By shifting the focus from more regulation to more investment and empowerment, we can create an environment where American companies can compete and win — from the frontier of the cloud to the edge of the battlefield.

Learn more.