Вход на сайт

Просмотр новости

Найдите то, что Вас интересует

US Army: Experimental AI Did Not Influence $449M White Sands Contract Award

Дата публикации: 06-08-2026 19:14:18

Disputed Army procurement offers unprecedented public view into how officials tested AI during proposal reviews and how that experiment became evidence in court.

Основное содержимое страницы с новостью.

Court filings publicly released July 31 in a bid protest over a $449.4-million U.S. Army mission support services contract identify an internal artificial intelligence platform tested during the source-selection process and describe why Army officials say it did not influence the award.

The filings, originally submitted July 10 and later released in redacted form, provide the government's first detailed public account of how an internal Army artificial intelligence platform known as FAST TRACK entered the procurement process.

TRAX International Corp., the unsuccessful bidder on the contract, alleges that unsupported evaluation findings resembling AI-generated errors contributed to the Army’s selection of incumbent Southwest Range Services LLC. The Army maintains that the formal evaluation was completed by human reviewers and that FAST TRACK’s output played no role.

Read More Trax Int’t Corp. v. The United States, et al. —Bid Protest, Cross-Motion for Judgment

The cost-plus-fixed-fee contract covers engineering and technical support for White Sands Test Center in New Mexico, including experimentation, research, weapons and systems testing, range operations and training. It is not a construction or civil engineering contract.

The Army awarded Southwest Range a three-month phase-in period, nine-month base period, four one-year options and a possible six-month extension.

Southwest Range’s evaluated price was $449.4 million, about $29.4 million, or 6.5%, above TRAX’s $420-million proposal. Both received outstanding overall mission-capability ratings, but Southwest Range was rated outstanding for continuity of operations while TRAX was rated good. The Army determined that Southwest Range’s additional technical strengths justified the premium.


Army Says AI Was a Separate Test

According to a partially redacted July 10 court declaration from a procurement analyst with the Army's Mission and Installation Contracting Command, Army Contracting Command-Detroit demonstrated an internal AI platform known as FAST TRACK on May 1, 2025, as officials considered whether the technology might support future source selections.

Looking for quick answers on construction and engineering topics?
Try Ask ENR, our new smart AI search tool.
Ask ENR

The analyst said leadership directed the evaluation of FAST TRACK after it was initially tested using fictitious acquisition documents.

Read More The United States, et al. —Procurement Analyst Declaration (Redacted)

According to the declaration, the analyst selected the White Sands procurement and another active competition at Dugway Proving Ground, Utah, because both were underway, but received "clear instructions that none of the results generated from the tool would be used to inform the source selection evaluation boards for any real-world acquisition."

The exercise, the declaration states, was intended to assess the platform's capabilities rather than serve as part of the White Sands procurement.

Read More The United States, et al. —White Sands Source Selection Evaluation Board Declaration (Redacted)

According to a separate declaration from the White Sands Source Selection Evaluation Board chair, the four-member board completed its review of the first proposal, identified as Offeror A, before a procurement analyst projected FAST TRACK’s analysis on a screen. Board members did not receive electronic or paper copies of the output, the chair said.

The declaration states the board found the AI results unusable and did not revise its assessment of Offeror A. It did not consult AI while evaluating the other two proposals, and any errors in the completed evaluation would have resulted from human review rather than the platform.


Bid Protester Challenges Army's Account

TRAX's complaint offers a different version of the story. The company alleges the Army identified a weakness in its proposal based on references that did not appear in the proposal and argues the error bears the hallmarks of an AI "hallucination."

Read More Trax Int’t Corp. v. The United States, et al. —Response in Opposition

The complaint also questions whether some favorable findings credited to Southwest Range originated with human evaluators or the experimental AI analysis.

For its part, the Army acknowledged during an earlier U.S. Government Accountability Office protest that the weakness assigned to TRAX was erroneous. GAO nevertheless denied the protest May 14, finding that the error did not competitively prejudice TRAX because the source-selection authority’s comparison relied on Southwest Range’s technical advantages rather than the unsupported weakness.

The U.S. Justice Department subsequently opposed TRAX's request to add the AI materials to the court record. The government argued the disputed output was not part of the source-selection record because evaluators did not rely on it. As a result, it contended that adding the materials would supplement, rather than complete, the administrative record.


Court Orders AI Analyses Produced

U.S. Court of Federal Claims Judge Carolyn N. Lerner granted TRAX's request and ordered the government to produce FAST TRACK analyses associated with all three proposals and place them in the administrative record. She also directed the parties to submit briefs addressing the relevance and potential effect of the additional documents.

An Air Force Research Laboratory computer scientist works on NIPRGPT, an Air Force-developed generative AI platform.

An Air Force Research Laboratory computer scientist works on NIPRGPT, an Air Force-developed generative AI platform. The image illustrates the Defense Department's broader development of AI tools; the White Sands bid protest involves a separate internal Army platform known as FAST TRACK.

Photo: Albert P. Santacroce/USAASC

The judge's order is not offering an opinion on whether AI impacted the award, rather it is simply entering the relevant materials into the court's record. Many of the AI analyses and supporting declarations remain under seal.

Southwest Range's July 31 filing notes that although TRAX alleged AI hallucinations in its complaint, it did not pursue those claims in its original motion for judgment on the administrative record. The joint venture said it would respond if TRAX renewed the allegations after the court expanded the administrative record.

The filing instead focuses on defending the Army's evaluation. Southwest Range argues its incumbent experience, management structure, workforce-retention plan and continuity systems supported the technical advantages identified by evaluators and justified the roughly $29.4 million price premium. The joint venture also argues TRAX cannot show it would have had a substantial chance of receiving the award even if the challenged findings were removed.

Beyond the AI allegations, TRAX also challenges Southwest Range's accounting system, the Army's consideration of past performance by joint venture members and the agency's investigation of an alleged organizational conflict of interest. GAO previously rejected those protest grounds or found that any errors did not competitively prejudice TRAX.

The case remains pending before the U.S. Court of Federal Claims. All three parties—TRAX, the United States and Southwest Range—have filed motions for judgment on the administrative record. Judge Lerner has not yet ruled on the merits.

The Army did not immediately respond to an ENR request for comment.

Схожие новости

#Наименование новостиТональностьИнформативностьДата публикации
1Contractor alleges Army inappropriately used AI to make $450M contract award09.8805-08-2026
2The Private Firms Powering China’s Military AI Push0524-03-2026
3 Pete Hegseth wages war on Anthropic -5624-02-2026
4Эксперт: США испытали ракету из установки, якобы не обладавшей наступательным потенциалом0021-08-2019
5 Donald Trump’s AI regime is opaque, unpredictable—and unsustainable -5630-06-2026
6Beyond the Hype: How AI Can Reshape Government Operations01028-05-2026
7Pentagon ready to deploy Salesforce AI agents for admin tasks07.0307-08-2026
8A new era in testing: USAF TPS partners with Stanford, Silicon Valley for AI, emerging technologies course5711-02-2025
9Собиравший досье на военные предприятия России попал на видео07.931-07-2026

Классификация: Наука. Схожих патентов: 0. Схожих новостей: 9. Тональность: 0. Информативность: 8.36. Источник: www.enr.com.