A federal lawsuit alleges that staff from the Department of Government Efficiency (DOGE), a small-agencies review team embedded at the General Services Administration, used the AI chatbot ChatGPT and keyword searches to identify and recommend cancellation of a large number of National Endowment for the Humanities (NEH) grants because they were judged to relate to diversity, equity, and inclusion (DEI).

According to court filings and deposition testimony from DOGE staffers Nathan Cavanaugh and Justin Fox, reviewers scanned spreadsheets of short grant summaries and searched federal grant records for keywords such as “gay,” “BIPOC,” “indigenous,” “tribal,” “melting pot,” and “equality.” Fox testified that he ran hundreds of grant descriptions through ChatGPT and asked whether each description related to DEI without providing the system a definition of DEI. The plaintiffs’ filings say the AI flagged many projects as DEI-related because they referenced marginalized communities or topics involving race, religion, gender, or sexuality. Cavanaugh and Fox acknowledged they lacked academic or humanities backgrounds, did not consult the NEH peer-review process or outside subject-matter experts, and based judgments on reading summaries.

The plaintiffs — including the Modern Language Association, the American Council of Learned Societies, and the American Historical Association — allege the DOGE review process led to termination notices for more than 1,400 active NEH grants totaling over $100,000,000, representing roughly 97 percent of the agency’s active grants, and that some notices were sent outside NEH’s normal grants system using an unofficial Microsoft email address with near-identical wording and no individualized explanations. Internal records cited in filings include an acting NEH chair warning that many cancellations lacked justification while also noting that final authority on cancellations rested with the DOGE team.

Specific grant impacts cited in filings and depositions include: a $349,000 NEH grant to the High Point Museum in North Carolina for HVAC replacement to preserve collections and increase public access that was canceled after an entry in a spreadsheet showed ChatGPT classifying the project as related to DEI; Colorado projects reviewed via the AI process, including a History Colorado boarding-school oral-history project that retained full funding, a University of Colorado Boulder musical-heritage project (Soundscapes of the People) and a curriculum project on Indigenous languages and cultures that were flagged and lost remaining federal funding, and a History Colorado podcast (Lost Highways) that was reviewed but not flagged; projects on LGBTQ military service, feminist and queer perspectives on HIV/AIDS activism and prison abolition, research on the 1873 Colfax Massacre, a biography of jurist Oscar Adams Jr., translations of Holocaust-era Soviet Jewish fiction, a Holocaust documentary about Jewish women forced into labor, archival work on Italian American communities, and other projects addressing Black civil-rights history. Depositions say some projects were flagged solely because they used terms such as “LGBTQ,” “homosexual,” or “Tribal.”

Testimony quoted in filings indicates reviewers sometimes relied on personal judgment rather than scholarly sources; one reviewer acknowledged not consulting books or other academic materials to guide decisions. Reviewers testified that their actions were guided in part by an executive order addressing diversity programs, but could not always define DEI or recall the order’s contents. Reported pay for the reviewers during DOGE work was cited in filings as $150,000 for Fox and $120,000 for Cavanaugh.

The lawsuit challenges the broader set of NEH grant cancellations as unlawful and alleges the administration replaced expert review with a rushed system that relied on an AI chatbot. Plaintiffs contend the terminations harmed scholarship, cultural preservation, and public access to materials. DOGE staff quoted in filings said reducing government spending and the federal deficit informed their actions. The litigation is ongoing; court filings, depositions, and internal NEH communications cited in the case remain part of the record.

Original Sources: 1, 2, 3, 4, 5, 6, 7, 8 (doge) (chatgpt) (dei)

Actionable information: The article describes a lawsuit and alleges that a government review unit used ChatGPT to flag and cancel NEH grants tied to DEI, naming a specific cancelled $349,000 grant and describing how a spreadsheet of the chatbot’s classifications guided cancellations. That reporting is news but it gives almost no practical, step‑by‑step actions an ordinary reader can take right now. It does not offer templates, forms, or clear instructions for affected institutions, grant applicants, or members of the public. The closest actionable element is identification of the parties involved (NEH, DOGE, plaintiffs such as the American Council of Learned Societies and the American Historical Association) and the fact of litigation, which suggests legal avenues exist, but the article does not explain how a reader could join a legal challenge, appeal a cancellation, or access recovery clauses. In short: the article reports what happened but offers no concrete steps a reader can follow immediately.

Educational depth: The piece conveys factual detail about who was involved, what grant was canceled, and that ChatGPT was used to classify proposals, but it stays at the descriptive level. It does not explain how the chatbot was prompted, what classification criteria or prompts were used, the confidence or error rates of the AI, or whether human review supplemented the tool. It does not analyze the legal standards at issue, such as the specific statutory or constitutional arguments being advanced, how administrative procedure law might apply, or precedents about use of automated tools in government decision making. Numbers shown (the $349,000 grant and roughly 70 percent recovery) are reported but not contextualized: the article doesn’t explain whether that level of recovery is typical under termination clauses, what proportion of NEH funding was affected, or how many projects were flagged. Overall it informs about events but does not teach the underlying systems, processes, or reasoning in enough depth for a reader to understand causes or likely consequences.

Personal relevance: For most readers this is an informative news item about federal grant administration and AI use. It is directly relevant to a narrow group: applicants to NEH or similar grantmakers, cultural institutions dependent on such grants, and scholars concerned about public funding for the humanities. For the general public the relevance is more indirect—it raises questions about government use of AI and administrative transparency—so the practical impact on an average person’s safety, finances, or day‑to‑day decisions is limited.

Public service function: The article serves a public interest by alerting readers to potential misuse of AI in government decision making and to litigation that could affect public funding for humanities projects. However, it offers little guidance for people who need to respond (grant applicants, nonprofit administrators, or concerned citizens). It does not provide contact information for affected parties, ways to monitor the litigation, or instructions for submitting public comments or petitions. As public service journalism it flags an issue but stops short of providing tools the public could use to act responsibly or protect their interests.

Practical advice quality: There is essentially no practical advice in the article. It reports what specific actors alleged and what happened to one grant but does not offer realistic steps for grant recipients facing cancellation, institutions drafting contracts, or funders implementing review processes. Any reader seeking to protect a project or challenge a cancellation would need legal and administrative guidance not provided here.

Long‑term impact: The article could be a starting point for discussion about the long‑term implications of relying on AI in government decisions, and the lawsuit could set precedents. But the article itself does not provide frameworks or recommendations for institutions to plan ahead, build resilience into funding agreements, or adapt review processes. Its focus is on a discrete incident and litigation, so it offers limited help for longer‑term planning.

Emotional and psychological impact: The reporting might provoke concern or frustration, especially among people in the humanities or nonprofit sectors. Because it mainly recounts allegations without offering constructive next steps or context about remedies and safeguards, it may leave readers feeling unsettled or helpless rather than better informed about options.

Clickbait or sensationalizing: The article does not rely on obvious clickbait phrasing in your summary. It reports a striking allegation—government use of ChatGPT to cancel grants—and names a concrete amount and institution, which is attention‑grabbing. The coverage is newsworthy, but the piece leans on the shock value of an AI tool being used in grant cancellations without adding deeper explanation, which amplifies sensational impact without substantive follow‑up.

Missed opportunities to teach or guide: The article missed several chances. It could have explained how government grant reviews normally work, what constraints govern the use of automated tools by agencies, common contractual protections (like termination clauses and how recovery typically works), or practical steps applicants can take to reduce risk from sudden funding changes. It also could have described how to evaluate an AI classification (questions about prompts, training data, human oversight, audit trails) and what transparency or documentation to demand when public agencies use algorithms.

Concrete, practical guidance the article failed to provide

If you are an applicant or nonprofit that relies on government grants, review your grant agreements early for termination and recovery clauses so you understand remedies and allowable pre‑award expenditures. Keep careful records of work done and costs incurred before an award is finalized to support claims for reimbursement under termination provisions if a grant is canceled. Maintain copies of proposals, correspondence, and receipts in a single organized folder so you can respond quickly to disputes.

If you’re monitoring agency use of AI, ask for and keep requests in writing. When interacting with a government office, request specific explanations of decisions and whether automated tools were used. A written request for the rationale and any documents the agency relied on creates a paper trail useful for internal appeals or litigation later.

If you’re a board member or manager at a cultural institution, build contingency plans that assume some grants may be delayed or cut. Prioritize projects so those with the highest risk to collections or safety have alternative funding or phased implementation. Consider establishing a small reserve or flexible fundraising plan that can be activated when grant funding is uncertain.

If you’re concerned about public policy or accountability, compare independent accounts before drawing conclusions. Track filings in the lawsuit or public records releases to see the agency’s stated procedures and any audit or report that follows. Encourage transparency by contacting elected representatives or oversight bodies with concise, fact‑based requests for information about how automated tools are authorized and audited.

If you are an individual reader seeking to make sense of similar stories, look for three things to assess reliability: named sources and documents (such as court filings), concrete figures or documents rather than vague claims, and follow‑up coverage that shows responses from the agency involved. These elements help separate isolated allegations from established facts.

These are general, practical steps you can use to protect projects, demand transparency, and interpret similar stories more effectively without relying on outside databases or specific legal advice.

"used the AI chatbot ChatGPT to identify and cancel grants tied to diversity, equity, and inclusion programs" This phrase frames the tool as the main agent doing cancellations. It makes readers see ChatGPT as the actor, not the people who used it. That shifts blame from human decision-makers to a machine and hides who made the choices. It helps the view that AI caused harm rather than people who directed it.

"recorded the chatbot’s classifications and explanations in a spreadsheet, and used that spreadsheet to guide which grants to cancel." This wording compresses actions into a tidy, causal chain that makes the process sound mechanical and decisive. It hides nuance about other inputs or human judgment by implying the spreadsheet alone guided cancellations. That favors a narrative of automated, impersonal decision-making.

"One entry in the spreadsheet showed ChatGPT classifying the High Point project as related to DEI on the basis that better preservation could support greater access for diverse audiences." This quote uses a concrete example to make the classification seem absurd. By focusing on that single reasoning, the text invites readers to see the AI decision as trivial or wrong. It highlights a specific link to sound critical without showing how often or consistently this logic was applied.

"A DOJ deposition quoted in the filings identifies a DOGE staffer, Justin Fox, as saying employees used ChatGPT to analyze grant descriptions and determine DEI connections." Naming the staffer and citing a DOJ deposition gives the claim authority and makes it seem proven. The text presents this as fact without noting context or response, which leads readers to accept guilt or wrongdoing as established. That biases toward believing the described practice occurred exactly as stated.

"The American Council of Learned Societies and the American Historical Association brought the lawsuit ... contending that the administration replaced expert review with a rushed system that relied on an AI chatbot" This phrase uses strong contrast "replaced expert review" versus "rushed system" to paint the administration's action as reckless and anti-expert. It frames the story as experts versus a hasty, incompetent process, favoring the plaintiffs' perspective and undercutting any defense or mitigating details.

"contending that the cuts were unlawful and violated the First Amendment." The text reports the legal claim plainly, which is appropriate, but offers no counter-argument or government justification here. That leaves the reader with only one legal framing and can bias toward seeing the cancellations as likely illegal without showing opposing legal reasoning.

"Public statements from an academic organization characterized the process as showing disregard for the democratic process and for the value of the humanities." This strong language—"disregard for the democratic process"—casts the action as not only procedural but moral and political harm. The quote amplifies emotional judgment and broad social consequences without presenting the other side, steering readers toward moral condemnation.

"the institution began work before the award was terminated and later recovered roughly 70 percent of the grant through a termination clause." This detail emphasizes the museum's proactive work and partial recovery, which evokes sympathy and suggests harm. By highlighting the 70 percent recovery, the text both shows loss and mitigation; the way it's placed invites readers to view the institution as a victim despite partial restitution.

"Other projects flagged included a proposal from North Carolina Central University to develop teaching materials using digital archival collections." Listing another academic project as "flagged" reinforces a pattern that academic or educational work was targeted. The single example supports the narrative that scholarly programs were swept up, favoring concern for academia without showing a full sample or how many projects were truly affected.

"the Department of Government Efficiency, known as DOGE" Using the acronym "DOGE," which is also an internet meme, subtly invites readers to see the department as less serious or even foolish. That choice of shorthand creates a mocking tone that undermines the agency's authority and biases readers against it.

The text conveys anger and indignation through phrases like “alleges,” “used the AI chatbot… to identify and cancel grants,” and references to a “rushed system” and “disregard for the democratic process.” These words and the way the lawsuit is described signal frustration and moral outrage by those challenging the grant cancellations. The anger is moderate to strong: it is not shouted but is sustained across the passage by repeated mentions of allegedly improper procedures, canceled funding, and organizational complaints. This emotion serves to cast the defendants’ actions as unjust and to rally sympathy for the plaintiffs and affected institutions, encouraging the reader to view the cancellations as wrongful and worthy of challenge.

The passage also carries a clear sense of disappointment and loss, especially in the description of the High Point Museum’s situation: the museum “had sought the funds to replace an aging HVAC system to preserve its collections,” “began work before the award was terminated,” and “later recovered roughly 70 percent of the grant.” Words like “aging,” “preserve,” and “terminated” evoke the practical harm and setback experienced by the museum. The disappointment is moderate in intensity and is concrete in its details, which encourages the reader to feel sympathy for an institution harmed by administrative decisions and to understand the tangible consequences of the alleged actions.

There is anxiety and concern about process and fairness embodied in phrases such as “replaced expert review with a rushed system,” “relied on an AI chatbot,” and “the cuts were unlawful and violated the First Amendment.” These expressions introduce worry about institutional competence, legality, and constitutional rights. The concern is strong because it ties procedural shortcuts to potential violations of democratic and legal norms. This emotion aims to alarm the reader about broader implications beyond a single canceled grant, prompting attention to systemic risk and the need for scrutiny.

The text contains elements of distrust and skepticism toward the Department of Government Efficiency (DOGE) and its methods, shown by the reporting that DOGE “reviewed NEH grant proposals with the help of ChatGPT,” “recorded the chatbot’s classifications,” and “used that spreadsheet to guide which grants to cancel.” The naming of a staffer and the chain of actions builds a narrative that invites suspicion about intentionality and accountability. The distrust is moderate and works to undermine confidence in the agency’s competence and motives, nudging the reader to question whether the cancellations were fair or well-founded.

A subdued note of defensiveness and institutional pride appears in the way plaintiff organizations are presented: “The American Council of Learned Societies and the American Historical Association brought the lawsuit” and public statements “characterized the process as showing disregard for the democratic process and for the value of the humanities.” These formulations frame the academic organizations as protectors of professional standards and cultural value. The pride is mild but purposeful; it positions these groups as authoritative defenders of norms, encouraging the reader to lend credibility and support to their challenge.

The narrative quietly evokes moral urgency through phrases like “violated the First Amendment” and “disregard for the democratic process.” These strong moral terms heighten the stakes from administrative error to constitutional and civic harm. The urgency is significant enough to suggest that the matter is not merely bureaucratic but has wider civic ramifications. This emotion intends to mobilize readers’ ethical concerns and push toward scrutiny or action.

The emotional tone is shaped by word choice that favors charged or evaluative language over neutral phrasing. Instead of saying grants “were reviewed,” the text repeatedly uses verbs and nouns with stronger connotations—“used,” “cancel,” “terminated,” “rushed system,” “disregard”—which tilt the reader toward a critical interpretation. The repetition of the connection between ChatGPT and grant cancellations reinforces suspicion and makes that link seem central and decisive. Personal detail, such as naming the High Point Museum and giving the dollar amount and the museum’s specific need, humanizes the impact and increases emotional resonance by turning abstract policy into a concrete story of loss. Quoting a staffer’s reported admission and public statements from organizations adds authority and frames the narrative as contested between experts and officials. Together, these techniques intensify feelings of injustice, loss, and urgency, steering the reader to be sympathetic to the plaintiffs, skeptical of the agency’s methods, and concerned about broader procedural and constitutional consequences.