Washington’s emerging AI trusted partners system is drawing a sharp comparison with Ukraine’s battlefield technology model, where frontline units can access tools through published rules rather than opaque selection. Eli Lehrer, president and co-founder of the R Street Institute, argues that Ukraine’s war-driven openness offers a useful warning as the White House decides who gets early access to the most capable artificial intelligence systems.
Lehrer points to Ukraine’s drone war as the clearest example. Near the Donetsk front, Ukrainian drone teams can submit verified strike footage, receive points for destroying assigned targets and move up a public leaderboard. Higher scores help units obtain better equipment faster through an online marketplace that Ukrainian fighters have compared to Amazon, according to materials cited by Lehrer.
The procurement setup behind that system gives combat units direct access to drones from hundreds of manufacturers, many of them small Ukrainian shops. Lehrer says the process has cut purchasing timelines from months to days and can put new designs in the field within about a month of leaving a workbench. Feedback from soldiers can also reach manufacturers the same day, which makes the system less dependent on a central acquisition office.
What is the AI trusted partners plan?
The White House’s June executive order says a limited set of “trusted partners” should receive early access to advanced AI models, including systems that may find software flaws faster than human teams. The order does not define that phrase publicly, and Lehrer says the selection process runs through classified channels.
Advanced AI models, including large language models, work by predicting and generating outputs from patterns learned during training, which can make them useful for code analysis as well as ordinary text tasks. For a plain technical refresher, see Kernel’s guide to how LLMs work.
Lehrer’s complaint is not that the government wants safeguards around powerful models. His argument is that access rules matter. He says the administration has already used national-security and commerce authorities to restrict, suspend and then clear frontier models without publishing criteria that outsiders can inspect.
Ukraine has also opened threat intelligence more broadly. Its defense ministry recently launched TrophyLab, a platform that shares technical information about captured Russian weapons, including schematics, known weaknesses and physical samples, with allied militaries, intelligence services and hundreds of Ukrainian and partner-country companies. Lehrer says access is vetted, can be revoked and is governed by published criteria.
Why does Ukraine’s model matter for AI access?
Lehrer argues that Ukraine’s approach rests on three practical choices: distribute capability widely, avoid single points of failure and share threat information by rule. In his view, an undefined trusted-partner tier in the United States risks doing the opposite by concentrating the best tools among organizations that already have the strongest defenses.
The likely losers, he says, are not major banks or wealthy university hospitals. They are smaller institutions that may need help against AI-assisted cyber threats but may not qualify for early access under a closed system.
The United States already has sector-based centers for sharing threat intelligence, Lehrer notes. His argument is that Washington should make AI access standards public and legible, rather than letting access depend on discretionary decisions that affected defenders cannot see.
This story draws on original reporting from Techdirt.
