Industry Floats Export Control, Export Promotion Suggestions for AI Export Program
Advanced technology and AI companies largely supported the Commerce Department’s new effort to create a program aimed at increasing U.S. exports of AI technologies and services, with some saying companies should commit to "rigorous" export compliance conditions before being allowed to participate. One company said the U.S. should require businesses to automate their compliance for exports involving certain dual-use AI models, saying manual compliance presents too many “failure points.”
Sign up for a free preview to unlock the rest of this article
Export Compliance Daily combines U.S. export control news, foreign border import regulation and policy developments into a single daily information service that reliably informs its trade professional readers about important current issues affecting their operations.
The public comments, released by Commerce in recent days, came after the agency asked industry for feedback on how it should shape the AI exports program, including to ensure participants comply with export controls and other national security regulations (see 2510220008). The White House said the program’s goal is to spread the adoption of American AI systems around the world 2507240019).
U.S. companies and trade groups broadly applauded that mission in their comments to Commerce. Hydra Host, an AI chip infrastructure company, called the program a “critical national security and economic imperative,” while Americans for Prosperity, a political advocacy group, said the American technology industry would benefit “greatly” from new markets, adding that “embracing export opportunities is essential.”
Other firms urged Commerce not to limit the program to AI software and models. Boston Dynamics, an advanced robotics company, said the initiative should also promote exports of U.S.-made, dual-use robotics hardware, especially hardware that incorporates AI.
“Advanced robotic technologies consist of highly agile mobile robots designed to be put to work in a variety of environments and applications, and whose functionality is trained by, or enhanced by, artificial intelligence methodologies,” the company said. “These types of robots raise heightened national security risks and therefore should most squarely be the focus of export support policies.”
Several commenters said Commerce should require participants to meet a high export compliance threshold. Accrete, a company specializing in AI reasoning systems, said the agency should “require a comprehensive set of safeguards that address the full lifecycle of AI development, deployment, and use,” including that there be “full traceability” of all exported hardware, software, datasets and model components; documentation showing U.S. or “trusted-partner origin” for all sensitive parts; “rigorous end-user vetting” aligned with Bureau of Industry and Security regulations; requirements for partner nations to “maintain secure environments that meet export control standards”; and more.
Aside from all exports complying with the Export Administration Regulations and International Traffic in Arms Regulations, Accrete also said Commerce should apply outbound investment and “foreign access protections” to ensure “no foreign ownership or control … would violate outbound investment rules or create undue influence over sensitive AI components.” This should include prohibitions on certain foreign entities “acquiring proprietary reasoning systems, training datasets, or agentic capabilities without U.S. approval,” the company said.
Hydra recommended less specific export compliance conditions, saying the U.S. must “aggressively export AI infrastructure” to stop Chinese companies, including Huawei, from “establishing irreversible footholds in strategic markets.” To do this, Commerce should create an "AI Infrastructure Fast Track" export classification for items under the program while establishing a presumption of approval license review policy for “consortium exports to priority markets.”
The burden of proof for an export application should be for the “denial, not approval,” Hydra said.
Commerce should also condition exports on “exclusive American” chip usage for 20 years, the company said, and it should create a "Huawei Replacement Bonus" for “verified infrastructure swaps” of Huawei technology for American technology.
“The greatest national security risk is moving too slowly while China establishes facts on the ground,” Hydra said.
Hydra, which has an office in Miami, also said Commerce should create a new “export field office” in Miami to oversee AI exports to Latin American countries and beyond. “Miami's unique position as a gateway to Latin America and direct connection point to the Middle East makes it an ideal location for a new export field office to oversee and coordinate these efforts,” it said.
Several commenters outlined other nations Commerce should select as priority markets or trusted partners under the program. Hydra pointed to the Gulf Cooperation Council Nations of the United Arab Emirates, Saudi Arabia, Qatar, Kuwait and Oman as priority markets, while Sterling Computers mentioned countries with governments that are already parties to a formal U.S. alliance or national security agreement, including NATO, and countries in the Western Hemisphere that are major economic or national security partners but not in a formal U.S. alliance.
TekStart Group, a Canada-based firm that provides sales, marketing and business development services to high-tech start-ups, said Canada should be considered a “trusted partner.” Qualifying Canadian companies should show their “alignment with U.S. foreign policy goals” and national security "commitments," including by “demonstrating a shared commitment to decreasing international dependence on AI technologies developed by countries of concern.”
One commenter focused more on specific export regulations overseen by BIS. Rohan Sharma, founder of an AI compliance platform ConformityX and the CEO of Zenolabs.AI, said BIS should require exporters using certain AI models to automate their compliance.
Sharma said the current framework for monitoring dual-use AI foundation models “relies heavily on self-attestation and static threshold reporting,” which is "insufficient to capture the dynamic liquidity of AI model weights across jurisdictional boundaries.” Those “manual compliance checks often fail” to flag what Sharma said are “model weight” transfers that are “individually benign” but may violate export control compute thresholds when “aggregated by foreign adversaries,” leading to “critical failure points in the current regulatory posture.”
He said BIS should mandate “Dynamic [Export Control Classification] Scoring,” which would require exporters of foundation models to use automated tools that can link AI models directly to ECCN 4E091. That ECCN imposes license requirements for exports of certain technology for the development of advanced AI.
The U.S. also should require automated screening against the Office of Foreign Assets Control’s Specially Designated Nationals List, Sharma said. This would mandate “that all model access points (API or Weight Download) be gated by automated OFAC/Sanctions screening to prevent inadvertent transfer to restricted entities,” he said. “We cannot govern 2025 technology with 2015 paperwork.”