The letter is dated September 19, 2026 and addressed to Secretary of Defense Pete Hegseth and Director of National Intelligence Jay Clayton, with a copy to Platte B. Moring III, Inspector General of the Department of Defense. It is signed by three Democratic senators: Mark R. Warner, Vice Chairman of the Select Committee on Intelligence; Christopher A. Coons, Ranking Member of the Subcommittee on Defense; and Jack Reed, Ranking Member of the Committee on Armed Services. It asks the two officials to investigate two reported errors in AI-assisted military targeting and intelligence during 2026, and to give the relevant Inspectors General unrestricted access to both and to any unreported instances. It runs to three pages, five substantive paragraphs and four footnotes.
Summary of the letter
The two incidents. The letter opens on "public reports of an AI-targeting platform relying on outdated geospatial data in a kinetic strike in Minab, Iran" and "a public report on an aborted military interdiction operation driven by potential AI-hallucination in disseminated intelligence by U.S. Special Operations Command Pacific." Its first footnote cites the CNN report of September 18, 2026 by Katie Bo Lillis and Zachary Cohen on the second episode. The letter does not cite a source for the Minab account. Both are described at DOD — Department of Defense (AI Deployer).
The concern. The senators write that these events "have given us growing concern about the extent to which agencies under your oversight have prioritized acceleration of AI capability adoption and 'experimentation' over effective governance." They state that congressional oversight committees "await a more comprehensive accounting of these recent instances (including root cause analyses)," and that reports of AI targeting platforms having "repeatedly generated spurious outputs (based on outdated input data or outright hallucination)" risk "serious operational error" and undermine the confidence of the public and of international partners in US intelligence and "intelligence-informed kinetic actions."
The vendor reference. The second paragraph places the incidents "against the backdrop of the Department of Defense's harassment of a leading U.S. AI vendor over the latter's reluctance to allow its products' use in fully-autonomous kinetic operations." The letter does not name the vendor. Inside AI Policy, reporting the letter on September 24, identified the reference as the department's clash with Anthropic (Source: insideaipolicy.com); that identification is the outlet's, not the letter's. The dispute is described at Anthropic Claude Gov + Pentagon Dispute (2025–2026) and Anthropic v. United States (Pentagon ban challenge).
Intelligence tradecraft. The letter states that AI "has appropriate applications in analytics and other fields," but that "strong tradecraft standards should ensure that no disseminated intelligence product contains AI-generated errors, particularly when such intelligence may serve as the impetus for a kinetic action against a nuclear-armed adversary." It argues that "highly-publicized intelligence mistakes – potentially triggered by unnecessarily accelerating AI-capabilities ahead of effective testing, evaluation, and human oversight mechanisms – risks trivializing" two decades of progress by the Intelligence Community.
The governance record. The fourth paragraph states that the shortcomings of generative AI models, including "auditability, model hallucination (including mischaracterizations and false attribution), and lowered output reliability and accuracy-over-time," are well established in the technical literature. It then sets three earlier governance measures against the administration's record:
| Measure the letter cites | What the letter says of it |
|---|---|
| Public Law No. 119-60, 139 Stat. 718 (2025) (footnote 2) | Congress "passed legislation to establish stronger AI governance measures for contexts in which AI affected high-impact functions (such as targeting)," including oversight and compliance mechanisms for reviewing use and outputs, and appropriate human oversight. The letter does not give the act's name. |
| OMB M-25-21, April 3, 2025 (footnote 3) | "The Trump administration's own Office of Management and Budget in early 2025 had directed civilian agencies to apply these heightened measures to high-impact AI uses – while encouraging their usage in national security contexts." |
| National Security Memorandum 25 | Revoked by the administration; the letter describes it as "establishing similar AI governance requirements in national security contexts." Its replacement is NSPM-11. |
Against these, the letter states that the administration "has in the last year aggressively rolled back governance measures," citing the revocation of NSM-25 and the release of "a Department of Defense AI Strategy that treats AI governance policies as 'blockers' to AI adoption in military contexts." Footnote 4 cites that strategy as Hegseth's Artificial Intelligence Strategy for the Department of War of January 9, 2026, marking both "Sec'y of War" and "Department of War" with "[sic]" (Artificial Intelligence Strategy for the Department of War (2026)). The strategy's own text has a section headed "Wartime Approach to Blockers," which directs the department to "eliminate blockers to data sharing, Authorizations to Operate (ATOs), test and evaluation and certification, contracting, hiring and talent management, and other policies that inhibit rapid experimentation and fielding" (Artificial Intelligence Strategy for the Department of War (Hegseth memorandum, 2026)). The letter's phrase "AI governance policies" is its own characterization of that section. The letter continues: "A 'Move Fast and Break Things' approach may serve for high-engagement memes on social media platforms, but it risks potentially grave operational miscalculations and represents a serious threat to morale; recruitment; and foreign perceptions on the thresholds, accuracy, and justness of U.S. military action."
The request. The operative sentence reads: "We urge you to promptly investigate this matter, providing relevant Inspectors General with unrestricted access to both instances this year in which media reports have suggested significant errors in AI-enabled targeting workflows, as well as any additional instances that may have so far not been publicly reported." Earlier, the letter asks that the findings carry "a higher degree of public transparency." It sets no response deadline and poses no numbered questions.
Release and coverage
The letter bears the date September 19, 2026, a Saturday; CNN's report of it states, in a correction, that it was sent on Saturday (Source: cnn.com). Warner's and Reed's offices each issued a press release on September 21, 2026 linking the signed letter (Source: warner.senate.gov; Source: reed.senate.gov). Both releases give Coons's title as Ranking Member of the Senate Appropriations Committee Subcommittee on Defense, which the letter's signature block shortens to "Subcommittee on Defense." Warner's release quotes the tradecraft sentence as "the analytic and other fields"; the signed letter and Reed's release read "analytics and other fields." Inside AI Policy, which hosts the copy captured here, reported the letter on September 24, describing the senators as setting out potential oversight priorities should Democrats win a Senate majority in November (Source: insideaipolicy.com).
The letter follows a House letter of March 12, 2026 from Democrats led by Reps. Sara Jacobs, Yassamin Ansari and Jason Crow, which asked Hegseth whether AI, including the Maven Smart System, was used to identify the Minab school as a target (House Democrats' letter to Secretary Hegseth on civilian harm in Operation Epic Fury (2026)). The House letter asks ten questions of the department; the Senate letter asks for Inspector General investigation and extends the request to the Director of National Intelligence, whose remit covers the disseminated intelligence product in the SOCPAC episode.
No response from the Department of Defense or the Office of the Director of National Intelligence had been reported as of September 25, 2026.
Provenance
The raw file is a text-layer extraction of the three-page PDF posted by Inside AI Policy, captured on September 24, 2026. The same letter is linked from the senators' own sites at warner.senate.gov and reed.senate.gov, and Reed's release reproduces the full body text, which matches the capture. In the extraction the Coons signature image reads "Chris Coan" and Reed's title line appears after the copy block; neither affects the text of the letter. Footnote 3's reference "M25-21" is as printed.
Relationships
- supports: DOD — Department of Defense (AI Deployer) — a congressional request for Inspector General review of the Minab and SOCPAC episodes.
- related: Pete Hegseth, Jay Clayton — addressees.
- related: Mark Warner, Christopher Coons, Jack Reed — signatories.
- related: House Democrats' letter to Secretary Hegseth on civilian harm in Operation Epic Fury (2026) — the House letter on the Minab strike.
- contradicts: Artificial Intelligence Strategy for the Department of War (2026), NSPM-11 — Artificial Intelligence in the National Security Enterprise (Trump, June 5, 2026) — the letter opposes the strategy's treatment of "blockers" and the revocation of NSM-25.
- depends-on: OMB M-25-21 — Accelerating Federal Use of AI through Innovation, Governance, and Public Trust — cited as the administration's own high-impact AI requirements.
- related: Anthropic Claude Gov + Pentagon Dispute (2025–2026) — the vendor dispute the letter refers to without naming the company.
- related: Autonomous Weapons, Sycophancy and Hallucination, Project Maven (Algorithmic Warfare Cross-Functional Team).