Comment letter

October 18, 2026

Regulatory Secretariat Division
General Services Administration
Attn: FAR Case 2026-003
1800 F Street NW
Washington, DC 20405

RE: Comments on the September 18, 2026, Revolutionary FAR Overhaul (RFO) Proposed Rule regarding artificial intelligence and automated analytic tools in proposal evaluation, price reasonableness analysis, and contract file documentation

Dear Members of the Federal Acquisition Regulatory Council:

On behalf of GovSignals, Inc. (GovSignals), I submit these comments on the Federal Acquisition Regulatory Council's (FAR Council) proposed rule of September 18, 2026, which seeks to convert the proposed RFO and interim class deviations for FAR Parts 8, 12, 13, 15, 38, 44, and 51 into permanent regulation, pursuant to Executive Order (E.O.) 14275, Restoring Common Sense to Federal Procurement.¹ This comment addresses the lack of coverage in the current FAR or in any of the interim or proposed changes to the FAR regarding the use of artificial intelligence (AI) and automated analytic tools by Government personnel in proposal evaluation, price reasonableness analysis, and contract file documentation.² This use (and misuse) has been highlighted in recent public information and bid protest actions and there is an urgent need for basic coverage in the FAR on the use of AI in federal acquisitions.

GovSignals is a small business and a nontraditional defense contractor under 10 U.S.C. 3014.

Introduction

GovSignals builds AI-native acquisition software. The pre-award and post-award portions of our platforms are in production use at U.S. Special Operations Command (USSOCOM),³ DARPA, Defense Innovation Unit (DIU) and U.S. Air Force.⁴ We can supply cycle-time and workload data from those deployments as an evidentiary appendix without reference to any specific solicitation.

We support the RFO initiative and generally support the proposed rule, but for the gap in coverage for AI use in federal contracting. The proposed rules' preference for permissive, plain-language text and its reorganization of Part 15 around the acquisition lifecycle seeks to and will assist contracting officers. Our recommended changes align with the RFO's philosophy by extending the scope of the RFO to encompass language regarding AI. While we recognize and support the effort to streamline the FAR, and acknowledge the broad permissive authority granted to buying activities to use any technique appropriate, the FAR should still affirmatively provide that Government evaluators and contracting officers may use AI and automated analytic tools in solicitation design, source selection evaluation, and price analysis, and should, within the administrative record and contract file, articulate the process used and provide a detailed audit trail. We do not propose mandating any specific AI tool or process or removing judgment from any Government official.

Instead, the issue is that the current FAR and the proposed rule are silent regarding the authority of agencies that are already using these tools in live source selections. A contracting officer with a tool already deployed in their agency has to decide whether to forgo utilizing the tool or using it without a regulatory standard. If the contracting officer utilizes the tool, it may invite a protest such as the one TRAX International Corporation pursued against the Army over the White Sands Missile Range support services contract. Regarding TRAX, Army Contracting Command-Detroit ran its FAST TRACK AI platform during the active White Sands Missile Range competition, which resulted in a $449 million award to Southwest Range. TRAX alleged in its post-award bid protest at the Government Accountability Office (GAO) and at the Court of Federal Claims (COFC) that the tool produced a hallucinated weakness with fabricated references. GAO denied the protest in May 2026. COFC Judge Carolyn N. Lerner ordered that the AI tool's analyses be included in the agency's disclosed administrative record, even though TRAX's motion for judgment was denied.

If the buying activity uses an AI tool and does not disclose its use or subsequent analysis, the solicitation risks a pre-award protest and the award risks a post-award protest. If the contracting officer does not use the tool out of fear of protest or other reasons, then they forfeit the potential gains from the use of the tool that is available for them.

We propose the following four-part approach. We believe it is the only shape likely to survive both the Council's stated preference against prescription and post-TRAX scrutiny at the GAO and the Court of Federal Claims:

1. Permissive. AI tools "may" be used. Nothing mandates AI.
2. Disclosed. If any AI tools are used in evaluation, the request for proposals (RFP) explicitly states this disclosure, in the same way FAR 15.104-2 already requires disclosure of whether negotiations will occur.
3. Human-adopted. Every AI-assisted finding is independently reviewed and adopted by a Government evaluator before it is relied upon; the source selection decision remains the Source Selection Authority's (SSA) independent judgment, as in existing FAR 15.205.
4. Documented. The contract file identifies the tool, its scope of use, the reviewing official(s) by function or position, not by name, and whether any AI tool findings or recommendations were modified or rejected.

This framing encourages workforce enablement and protest-proofing. AI is not a substitute for contracting officers or evaluators, and nothing in this comment treats it as one. We recommend the Council adopt the text in Section II. If the Council declines to add regulatory text, Section IV recommends the same substance be added to the preamble to the final rule and also included in the non-regulatory guidance.

References throughout this comment to "existing text" refers to the language in the September 18, 2026, proposed rule, unless otherwise noted.

I. The Proposed Rule Is Silent on AI Tools Agencies Already Use

The proposed rule restructures Part 15 into five subparts, formalizes the "highest technically rated with a fair and reasonable price" approach at 15.102-3, restricts the lowest price technically acceptable approach for agencies other than the Department of War at 15.102-2(c), narrows "deficiency" at 15.001 to nonconformance with a material term of the RFP, and expands the advisory multistep process at 15.104-8. It reorganizes Part 12 around acquisition phases and reduces Part 13 to noncommercial simplified acquisitions that cross-reference Part 12 for solicitation and evaluation procedures. Neither the proposed rule nor its preamble mentions artificial intelligence, automated tools, or algorithmic assistance anywhere in the proposed rewrite of Parts 12, 13, or 15; in fact, these are not mentioned anywhere in the current FAR, the initial RFO or the implementing class deviations.

At the same time, agencies are already using such tools in live source selections. Army Contracting Command-Detroit demonstrated its internal FAST TRACK platform in May 2025 and then tested it against two active competitions, including the White Sands Missile Range mission support services competition.⁵ USSOCOM has published solicitation language stating that the Government "may employ AI as a tool to assist in the analysis and review of offeror proposals," for summarizing proposal content, checking compliance with solicitation requirements, and highlighting potential strengths, weaknesses, or risks, with all final evaluation judgments and source selection decisions reserved to Government personnel.⁶ GAO reported in April 2026 that agencies acquiring AI struggle to find technical experts to evaluate proposals and do not regularly collect lessons learned.⁷ This current environment regarding published language encouraging AI for acquisition yet the grey area regarding contracting officers' use produces a problematic gap for AI implementation within the procurement and acquisition process.

Office of Management and Budget (OMB) Memoranda M-25-21 and M-25-22 direct agencies to accelerate AI adoption and acquisition and set minimum risk management practices for "high-impact" uses.⁸ OMB Memorandum M-26-04 adds contractual requirements for procured large language models,⁹ but does not cover the use of AI tools in contract evaluations. Department of War Instruction 8430.01, effective September 8, 2026, treats AI-assisted software development as a force multiplier and sets procedures across the software lifecycle.¹⁰ The General Services Administration's (GSA) OneGov agreement with OpenAI, which took effect October 1, 2026, makes discounted, consumption-based access to ChatGPT models available across the Federal Government; OpenAI has extended the same offer to state, local, and tribal governments, a public-sector workforce it estimates at roughly 23 million.¹¹ The environment cross-agency demonstrates the momentum towards adopting AI technology at scale. However, without clear guidance regarding its use beyond concerted efforts to adopt the technology, evaluators have the tools but lack the requirements or appropriate guardrails necessary to use them. Therefore, the gap is not aligned with AI governance or a lack of government momentum, but rather it is the procedure itself.

We recommend: The FAR Council should acknowledge in the final rule that AI and automated analytic tools are in use in Government source selections and price analysis today, and should provide a permissive, governmentwide standard in Parts 12, 13, and 15 rather than leave each agency and each contracting officer to improvise one or ignore one altogether.

II. Proposed Text

For each proposed provision identified below, we identify the existing published proposed rule, set out our recommended insertion in bold, and explain the rationale for the change.

(a) 15.001 Definitions.

Existing text: Proposed 15.001 defines seven terms: clarification, competitive range, deficiency, negotiation, proposal modification, proposal revision, and weakness.

We propose inserting in the appropriate places:

"AI-assisted determination means any evaluation finding, rating, price analysis, or other source selection input generated in whole or in part by an artificial intelligence or automated analytic tool and relied upon by Government personnel in the acquisition."

"Artificial Intelligence" means the technologies and systems defined in section 9401(3) of title 15, United States Code.¹²

Rationale: While Part 2 (definitions) is the preferred location for adding these definitions, that Part closed for comment on July 23, 2026, so a part-specific definition in 15.001 is the best available vehicle.¹³ The definition is keyed to reliance by Government personnel. If a tool summarizes a proposal and the evaluator develops an independent finding, the tool has not produced an AI-assisted determination upon which reliance was placed. If it identifies a weakness that the evaluator adopts, it has.

(b) New 15.104-9 Artificial intelligence disclosure.

Existing text: None. Our proposed new section would follow 15.104-8 (Advisory multistep process). Existing 15.104-2 requires competitive RFPs to "advise offerors whether the Government intends to evaluate proposals and award a contract with or without conducting negotiations."

We recommend inserting:

"15.104-9 Artificial intelligence disclosure. If the agency intends to use artificial intelligence or automated analytic tools to assist in evaluating proposals, the competitive RFP must advise offerors that such tools may be used, identify the evaluation factors to which they may be applied, identify the portion(s) of the solicitation that will be evaluated, and state that all AI-assisted determinations will be independently reviewed and adopted by Government evaluators before being relied upon in the source selection decision. Use of such tools does not relieve the agency of the requirement at 15.201(b)(4) to evaluate proposals based solely on the factors and subfactors in the RFP."

Rationale: The addition would place AI disclosure alongside the existing requirement to disclose intended negotiations, an RFP-level notice the Council already accepts. A new paragraph (e) in 15.103 would be an alternative location.

USSOCOM already discloses intended AI use, but its solicitation also described an "output-only system" whose generated report data was not saved. This failure to save may not survive a post-TRAX disclosure requirement if there was use of the output, regardless of whether saved. The proposed 15.205 addition addresses that missing record.¹⁴

Corresponding solicitation language also should be added to 52.215-1 and, for commercial acquisitions, added to 52.212-1 or 52.212-2, as discussed in Section III.

(c) 15.201(b)(1) Evaluation team.

Existing text: The SSA must "[e]stablish an evaluation team tailored for the acquisition that includes appropriate contracting, legal, logistics, technical, and other expertise to ensure a comprehensive evaluation of proposals."

We recommend inserting at the end:

"The evaluation team may be supported by artificial intelligence or automated analytic tools, provided that responsibility for each evaluation finding remains with a Government evaluator."

Rationale: The tool serves solely as support for the evaluation team. A Government evaluator would remain accountable for each finding.

(d) New 15.202(a)(4) AI-assisted evaluation.

Existing text: Existing (a)(1) requires agencies to evaluate "based solely on the factors and subfactors in the RFP and document the evaluation." Existing (a)(3) requires the Government to "[m]aintain a record of oral presentations to document what the Government relied upon in making the source selection decision." Paragraph (a) ends at (3).

We recommend inserting a new Subparagraph (4):

"(4) AI-assisted evaluation. Government evaluators may use artificial intelligence or automated analytic tools to assist in reviewing proposals, identifying strengths, weaknesses, significant weaknesses, deficiencies, and risks, and drafting evaluation documentation. Before any AI-assisted determination is relied upon, a Government evaluator must independently review it and confirm that it is supported by the content of the proposal and consistent with the evaluation factors in the RFP. AI-assisted determinations that have not been so reviewed and adopted may not be used in the evaluation or the source selection decision."

Rationale: The paragraph follows the approach in (a)(3), that already requires a record of what the Government "relied upon" in oral presentations. Requiring a finding to be "supported by the content of the proposal" addresses the fabricated reference allegation in TRAX. The final sentence is the most prescriptive part of our proposed addition. If the Council declines to adopt it here, we recommend its inclusion in the Part 15 Practitioner Album.

Section 15.202(d) needs no change. It already provides that "[c]ontracting officers are responsible for evaluating and documenting the reasonableness of the awarded price in accordance with subpart 15.4," which keeps the price analysis text in subpart 15.4, where the techniques list lives (see 15.404-1(b) below).

(e) 15.205 Source selection decision.

Existing text: "[t]he source selection decision must represent the SSA's independent judgement and be based on a comparative assessment of proposals against all source selection criteria in the RFP"; the decision "must be documented and include the rationale for any business judgments and tradeoffs made or relied on by the SSA, including benefits associated with additional costs"; the documentation "does not need to quantify the tradeoffs that led to the decision."

We recommend inserting at the end:

"Where AI-assisted determinations were used in the evaluation, the source selection documentation must identify the tool or tools used, the evaluation factors to which they were applied, the Government evaluators who reviewed and adopted the resulting findings, and any AI-generated findings that were modified or rejected. The source selection decision may not rest solely on an AI-assisted determination."

Rationale: This is the audit-trail provision. It gives GAO and the CoFC a file to review and gives the agency a defense. It builds on existing 15.205 language ("independent judgement," "relied on by the SSA") rather than creating a new standard, and makes the treatment of AI-assisted findings explicit without changing the SSA's role.

(f) 15.404-1(b) Price analysis techniques.

Existing text: "[t]he Government may use various price analysis techniques and procedures to ensure a fair and reasonable price, including, but not limited to," a list beginning with comparison of proposed prices received in response to the RFP and comparison of proposed prices to historical prices paid.

We recommend inserting a new numbered technique:

"(x) Use of automated price analysis tools that compile and compare prices paid by the Government or others, commercial catalog and market pricing, competitive quotation history, and independent Government cost estimates, provided the contracting officer reviews the output, confirms the validity of the comparison in accordance with paragraph (b)(2) of this section, and documents the basis for the determination in the contract file."

Rationale: Price analysis offers the highest value and least controversial place to establish that automated analysis is an acceptable technique. The list is already non-exhaustive ("including, but not limited to"), so this is clarifying rather than expanding: a tool may assist with comparisons and checks without displacing the contracting officer's required determination.

(g) 15.408-2(a) Documenting the negotiation.

Existing text: It requires the contracting officer to "[d]ocument in the contract file the principal elements of the negotiated agreement" and lists the price negotiation memorandum's elements, beginning with the purpose of the negotiation, a description of the acquisition, and the name, position, and organization of each person representing the contractor and the Government.

We recommend inserting a new numbered element:

"(x) Identification of any artificial intelligence or automated analytic tool used in the price or cost analysis and the contracting officer's confirmation of review of its output."

Rationale: This would apply the same documentation standard to the price negotiation memorandum as to the source selection record under 15.205.

(h) New 12.203(c)(2)(iii) Automated tools.

Existing text: 12.203(c)(2) requires the contracting officer to "[e]nsure that quotations and offers can be evaluated in an efficient and minimally burdensome fashion," with (i) Quotations ("broad discretion in establishing how quotations will be evaluated" under 12.201-1) and (ii) Offers.

We recommend inserting a new (iii):

"(iii) Automated tools. Contracting officers may use artificial intelligence or automated analytic tools to assist in comparing quotations or offers against the evaluation factors. When such tools are used, follow the disclosure, review, and documentation requirements at 15.104-9, 15.202(a)(4), and 15.205, tailored to the simplicity of the acquisition."

Rationale: The cross-reference applies the Part 15 requirements to commercial acquisitions without creating a separate standard. That is consistent with Part 12's requirement for "efficient and minimally burdensome" evaluations and the use of cross-references rather than repeating provisions in each applicable section.

(i) 12.204(a) Price reasonableness.

Existing text: The contracting officer must "determine the price to be fair and reasonable" and, "[w]henever possible, base price reasonableness on competitive quotations or offers." A third sentence directs that "[f]or offers using the procedures in 12.201-2, see part 15 or 14, as applicable, for additional price documentation requirements."

We recommend inserting after the second sentence:

"Automated market research and price analysis tools that compile prices paid, commercial catalog and market pricing, and competitive quotation history are an acceptable means of supporting the price reasonableness determination, provided the contracting officer reviews the output and the contract file identifies the tool and the data sources and outputs relied upon."

Rationale: While Part 10 (market research) is the better location for this addition, it closed for comment on July 23, 2026.¹⁵ Within the scope of this proposed rule, the commercial price reasonableness determination in 12.204(a) is the best location for recognizing the use of automated market research and price analysis tools.

(j) 12.204(b)(1) Documentation.

Existing text: This requires the contract file to include "a written description of the procedures used in awarding the contract and the number of quotations received."

We recommend inserting before the period at the end:

"and, where artificial intelligence or automated analytic tools were used in the evaluation or price analysis, the identity of the tool, the output generated, and the contracting officer's confirmation of review".

Rationale: This clause would maintain the same file standard as 15.205 and 15.408-2 across the parts.

(k) 13.202 Evaluation, 13.203(a) Price reasonableness, and 13.203(b) Documentation.

Existing text: 13.202 directs contracting officers to "[f]ollow the procedures in 12.203 when establishing evaluation factors and a process to evaluate quotations." No change; it inherits proposed 12.203(c)(2)(iii) by cross-reference. We say so expressly here and ask that the preamble confirm it. 13.203(a) requires the contracting officer to "determine the price to be fair and reasonable" and, "[w]henever possible, base price reasonableness on competitive quotations".

We recommend inserting at the end of (a):

"Automated price analysis tools may be used to support the determination in accordance with 12.204(a)."

We also recommend inserting before the period at the end of (b):

"and, where artificial intelligence or automated analytic tools were used in the evaluation or price analysis, the identity of the tool and the contracting officer's confirmation of review".

Rationale: 13.203(b) requires the file to include "a written description of the procedures used in awarding the purchase order and the number of quotations received". The cross-references apply the requirements without repeating the text. The proposed 13.203(b) language is identical to the insertion at 12.204(b)(1).

III. Corresponding Solicitation Provisions in Part 52

The disclosure requirement at 15.104-9 needs corresponding language in 52.215-1, Instructions to Offerors, Competitive Acquisition, and, for commercial acquisitions, in 52.212-1 or 52.212-2. All three provisions are within the scope of this notice: the proposed rule revises 52.212-1 and 52.212-2 in full at amended instruction 9 (91 Fed. Reg. at 59,458) and 52.215-1 and 52.215-2 in full at amended instruction 15 (91 Fed. Reg. at 59,464). Our recommended text can therefore be proposed in this comment. We also believe that our recommended changes can be addressed in the September 23, 2026, proposed rule for Parts 14, 28, 36, and 52 (FAR Case 2026-010, 91 Fed. Reg. 59,534).

Recommendation: The FAR Council should add to 52.215-1 and to 52.212-1 or 52.212-2 a paragraph corresponding to 15.104-9, so that the offeror learns from the solicitation itself that AI or automated analytic tools may assist in evaluation, which factor(s) they may be applied to, and that Government evaluators will independently review every AI-assisted determination before adopting or rejecting it.

IV. Alternatives

If the Council declines regulatory text because the overhaul is intended to remove procedural requirements, we recommend two alternatives.

First, the FAR Companion and the Part 12 and Part 15 Practitioner Albums should include a section on AI-assisted evaluation and price analysis addressing disclosure, human review, and file documentation.¹⁶

Second, the preamble to the final rule should state expressly that nothing in Parts 12, 13, or 15 prohibits the use of artificial intelligence or automated analytic tools in evaluation or price analysis when used consistent with OMB Memorandum M-25-21 and documented in the contract file. A preamble statement plus an album section would achieve most of the practical objectives of our comments with far less resistance.

Recommendation: If the FAR Council declines to add regulatory text, it should (1) add a section on AI-assisted evaluation and price analysis to the FAR Companion and the Part 12 and Part 15 Practitioner Albums and (2) state in the preamble to the final rule that nothing in the revisions to Parts 12, 13, or 15 prohibits the disclosed, human-adopted, and documented use of AI or automated analytic tools in evaluation or price analysis consistent with OMB Memorandum M-25-21 and the documentation and disclosure requirements.

V. Conclusion

GovSignals supports the Council's effort to return the FAR to its statutory origins and make it easier for acquisition personnel and industry to use. AI-assisted evaluation needs the same clarity. A Government evaluator may use an AI tool but takes responsibility for owning the findings. The solicitation should disclose the possible use of AI, and the file should be a documentation of the audit trail.

An attachment provides all of our recommended changes to the proposed rule in tabular format.

Thank you for considering these comments. We would welcome the opportunity to work with the FAR Council, GSA's Office of Government-wide Policy, and OMB's Office of Federal Procurement Policy on a clear rule that lets acquisition personnel use these AI tools without yielding the human judgment and recordation that the procurement system requires.

I would be pleased to answer any questions.

Respectfully submitted,

Derek Hoyt, PMP
Chief Executive Officer
GovSignals, Inc.
169 Madison Ave, Ste 11051, New York, NY 10016
derek@govsignals.ai
(202) 415-8986
CAGE 9V6Y7 · UEI H5KNWGJ55NQ6

Attachment: Appendix A. Consolidated Proposed Insertions

Appendix A. Consolidated Proposed Insertions

Keyed to 91 Fed. Reg. 59,406, September 18, 2026.

Notes

1. Federal Acquisition Regulation: Revolutionary FAR Overhaul Parts 8, 12, 13, 15, 38, 44, and 51, FAR Case 2026-003, Docket No. FAR-2026-0003, 91 Fed. Reg. 59,406 (Sept. 18, 2026) (comments due Oct. 19, 2026); Executive Order 14275, Restoring Common Sense to Federal Procurement (Apr. 15, 2025). Back
2. We acknowledge that Executive Order 14434 of September 29, 2026, directs federal agencies to begin using the term "Super Intelligence" but we have not done so in this comment letter. Back
3. U.S. Special Operations Command, SOF RACER, Other Transaction Agreement H92419269E018 (consortium management under 10 U.S.C. 4022). Back
4. Defense Innovation Unit, Enterprise Workflow and Reporting Platform (EWARP), Other Transaction Agreement HQ08452590047. Back
5. Bryan Gottlieb, US Army: Experimental AI Did Not Influence $449M White Sands Contract Award, Engineering News-Record (Aug. 6, 2026) (summarizing the July 10, 2026, declarations of the Mission and Installation Contracting Command procurement analyst and the Source Selection Evaluation Board chair, released in redacted form July 31, 2026). Back
6. U.S. Special Operations Command, Notice ID H9240026RE001 (SAM.gov), as quoted in Michael J. Brewer, Should the Government Use AI Tools to Evaluate Contract Proposals?, Burr & Forman Government Contracting Blog (Apr. 7, 2026). Back
7. U.S. Government Accountability Office, Artificial Intelligence Acquisitions: Agencies Should Collect and Apply Lessons Learned to Improve Future Procurements, GAO-26-107859 (Apr. 13, 2026). Back
8. Office of Management and Budget, Memorandum M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust (Apr. 3, 2025); Memorandum M-25-22, Driving Efficient Acquisition of Artificial Intelligence in Government (Apr. 3, 2025). Back
9. Office of Management and Budget, Memorandum M-26-04, Increasing Public Trust in Artificial Intelligence Through Unbiased AI Principles (Dec. 11, 2025) (contractual requirements for procured large language models). Back
10. Department of War Instruction 8430.01, Accelerated Mission Software (approved Aug. 31, 2026; effective Sept. 8, 2026). Back
11. General Services Administration, GSA Expands OneGov AI Offerings with Discounted, Consumption-Based Access to OpenAI's ChatGPT (news release, Sept. 10, 2026) (27-month agreement expected to take effect Oct. 1, 2026, with a 50 percent discount on consumption-based pricing). The figure of roughly 23 million eligible public-sector employees is OpenAI's estimate as reported by FedScoop (Sept. 10, 2026). Back
12. Derived from Section 3 of Executive Order 14434 of September 29, 2026. Back
13. Federal Acquisition Regulation: Revolutionary Federal Acquisition Regulation Overhaul Parts 1, 2, 4, 33, 39, 40, and 53, FAR Case 2026-001, Docket No. FAR-2026-0001, 91 Fed. Reg. 37,550 (June 23, 2026) (comments closed July 23, 2026); Federal Acquisition Regulation: Revolutionary Federal Acquisition Regulation Overhaul Parts 6, 7, 10, 18, 26, 37, and 41, FAR Case 2026-002, Docket No. FAR-2026-0002, 91 Fed. Reg. 37,636 (June 23, 2026) (comments closed July 23, 2026). Back
14. See supra, note 5. Back
15. See supra at note 13. Back
16. 91 Fed. Reg. at 59,407 (describing the FAR Companion as a non-regulatory resource providing insight from experienced practitioners); Practitioner Albums for the overhauled FAR parts are published at acquisition.gov/far-overhaul. Back

← Back to the press release