NEWSGUARD

Institutional Archetype

THE CREDIT BUREAU — NewsGuard is a private company that assigns a number to a news outlet the way a credit bureau assigns a score to a borrower, and the number travels. It does not host content, moderate a platform, or take anything down. It rates: nine “journalistic criteria,” a 0–100 reliability score, a color. The score is then licensed downstream — to advertisers deciding where their money goes, to platforms, and to AI companies training and grounding their models — so that a low rating functions, in effect, as a demonetization signal issued by a third party the rated outlet never hired. This profile scores that position — the ratings seat and where its output flows — not the accuracy of any individual rating, and not the motives of the people who assign them.

Mandate & Origin

NewsGuard Technologies was founded in 2018 and is run by co-CEOs Steven Brill and L. Gordon Crovitz, described on its own site as “veteran journalists and news entrepreneurs” — Brill the founder of The American Lawyer and Court TV, Crovitz a former publisher of The Wall Street Journal; the two previously co-founded Press+, sold to RR Donnelley.

Its self-stated mandate is to counter misinformation by “rating the credibility” of news and information sources using trained analysts rather than an algorithm — the company markets itself as “the opposite of an algorithm: Transparent and accountable.” The product line extends the score into commercial and government channels: BrandGuard (advertiser brand-safety), platform and browser integrations, and licensing to AI developers.

Funding & Backers

NewsGuard is a for-profit company, and its revenue mixes commercial licensing with government contracting. The government line is the documented one: on 7 September 2021 the U.S. Department of Defense awarded NewsGuard $749,387 (Air Force / U.S. Cyber Command), recorded on USASpending.gov, for work NewsGuard describes under its “Misinformation Fingerprints” program; NewsGuard’s own press release announced it had won a Pentagon/State Department contest for detecting COVID-19 misinformation. It is also listed on the portfolio page of SAIF, the Safe Artificial Intelligence Fund founded by Geoff Ralston, as “Global Leader in Information Reliability” (SAIF, portfolio); the instrument and date are not stated.

NewsGuard’s response to the funding criticism is on the record and is reproduced here in its own terms: co-CEO Crovitz has stated the Pentagon work “is focused on the identification and analysis of information operations targeting the U.S. and its allies,” that “our Pentagon contract is a single-digit percent of our revenues,” and that the company is not influenced by the government.

The cap table and the revenue line, from the company’s own page. NewsGuard’s investors page says that “Co-CEOs, Co-Editors-In-Chief and investors Steven Brill and Gordon Crovitz have governing rights in NewsGuard and control the editorial products of NewsGuard,” and that “Outside investors play no role in the determinations of ratings or the Nutrition Label write ups of websites.” It lists the outside investors “in descending order of their ownership interest”: 1898312 Ontario Inc. (Mark Simon), Cox Investment Holdings, Inc., the John S. & James L. Knight Foundation, Inc., Diatoms, Inc., Pharus Securities LLC, and the Safe Artificial Intelligence Fund LP (Geoff Ralston). On revenue: “NewsGuard’s revenue comes from consumer subscriptions to NewsGuard’s browser extension and Reality Check newsletter and from businesses licensing NewsGuard’s data” (NewsGuard, Our Investors).

The federal ledger, from the government’s side. USAspending, queried for recipients named “NewsGuard” across all federal award types on 23 October 2026, returns one record, the Air Force contract FA864921P1569, “MISINFORMATION FINGERPRINTS,” $749,387, 7 September 2021 to 8 December 2022, and no grant (USAspending). The House Oversight Committee’s release of 25 October 2024 adds that NewsGuard “revealed in a briefing with Oversight Committee staff that it has also performed work with the Joint Research Centre of the European Union, and the Global Engagement Center of the State Department, and with the Cyber National Mission Force within U.S. Cyber Command” (House Oversight Committee).

Institutional Voice & Intent

The register is journalistic and defensive: a company that calls itself a news organization, answers criticism with its ratings of its critics’ rivals, and describes every product as transparency. Its co-CEO Steven Brill, in the account on NewsGuard’s own site, put the design against the alternatives: “we would not tell any platform to block anything,” and “the worst would be to have the government make judgments about the reliability of what people were seeing online” (NewsGuard, Our Story). The free-speech page says the answer “is not to censor any content” but “to arm users with accurate information that helps them make their own decisions about whom to trust and what to believe” (NewsGuard, Free Speech and Censors). The company’s own account of the FTC fight is in its first-person statement: “an unprecedented year-long campaign to use government power to prevent NewsGuard from producing this journalism” (NewsGuard, Why We’ve Just Sued the Government). The AI turn is in its announcement of 12 May 2026: the browser extension now works “inside ChatGPT and Google Gemini — displaying its nonpartisan trust scores and providing access to its ‘Nutrition Labels’ next to every source those AI tools cite,” and COO Matt Skibinski says, “citing a source is only useful if you know whether to trust it” (NewsGuard press release).

Stated alignment: “An internet trust service that keeps users, brands, and democracies safe from misinformation online,” through transparent, journalist-made ratings and no fee from the sites it rates. Observed alignment: A licensed ratings feed sold to advertisers, platforms, AI developers and a defense customer, now also a chatbot, an AI-audit service and a plaintiff against a federal agency. Gap assessment: The company’s account is that it blocks nothing and licenses a label. The record shows the label is sold into the three places where a label gates money or sourcing, and that a federal agency treated the label as a boycott instrument. The file scores that position and not any rating.

Stated intent: Work with “technology platforms, news aggregators, internet service providers, advertising companies, major brands, educational institutions, governments, and researchers” using “human intelligence” (NewsGuard, About), and “arm users with accurate information” (NewsGuard, Free Speech and Censors).

Observed intent: Make the rating the default source-reliability signal across human browsers, advertising and AI. The company licenses a “FAILSafe” feed to AI developers, sells brand-safety data to advertisers, and launched its own chatbot in June 2026 whose responses are “drawn exclusively from 12,000 news and information sources rigorously vetted by NewsGuard journalists,” with revenue shared “50-50 with all news publishers whose journalism is cited” (NewsGuard AI). Interest, as documented: in its 2018 lobbying registration NewsGuard described the work as “Education and promotion of NewsGuards ‘human intelligence’ system of rating news and information websites and giving them ‘Nutrition Labels,’ in the hope that it will be seen by Congress and Internet platform companies as a logical middle course between government regulation to prevent false information on the internet or the platform companies continuing to attempt to fix the problem internally” (LDA registration, Prime Policy Group, 2018). Since the second quarter of 2021 an outside firm, ACG Advocacy, has registered for it on “Efforts to address misinformation online. Section 230 Reform.” (LDA registration, ACG Advocacy, 2021). A rule that put duties on platforms for what they carry would be a rule under which a licensed source-reliability feed has buyers; the filing says what the company hoped for and does not say what it would earn, and the file asserts no motive. The company’s own best case is the one above: of the three ways to rate news, a transparent, contestable human label is better than a government judgment or an opaque algorithm. CCDH and ADL both argued for Section 230 changes the same year, and the convergence is the finding.

Gap: NewsGuard’s defense is its transparency: published criteria, a published complaints policy, a revenue page and an investors page that names the owners. Its critics’ charge is the one in the FTC’s demand and the House letters, that the label functions as a demonetization signal. The record cannot settle whether any rating was wrong, and the file asserts none. For the chokepoint, it never needs to.

• Social accounts and feeds. X @NewsGuardRating, Bluesky, Facebook, LinkedIn and the Reality Check newsletter on Substack, all linked from newsguardtech.com; feeds newsguardtech.com/feed and newsguardtech.substack.com/feed (both HTTP 200, RSS).

Position in the Apparatus

NewsGuard sits where a source-reliability number meets four buyers.

• The advertiser feed and the order against it. The ratings serve as brand-safety data (the existing section above). NewsGuard says that in a merger review the FTC “entered an order requiring that NewsGuard be put on an unprecedented speech-based blacklist— i.e. the merged company had to agree not to subscribe to or rely on NewsGuard’s ratings and journalism,” and that the demand and the order came “at the urging of Newsmax” (NewsGuard, Why We’ve Just Sued the Government). The FTC’s 17 April 2026 filing in the case attached stipulated orders for Dentsu, WPP and Publicis (CourtListener docket).
• The AI feed. Beyond the ChatGPT and Gemini overlay, the company licenses “FAILSafe,” “created in response to a groundbreaking NewsGuard audit that found Russian website networks had infected top AI tools,” and publishes a monthly AI False Claims Monitor whose headline is that the 10 leading chatbots repeat false claims “28.79 percent of the time” (NewsGuard, FAILSafe for AI; press release, above). Its AI Tracking Center counts 3,749 AI content-farm sites in 16 languages, last updated 23 June 2026 (NewsGuard AI Tracking Center). A private ratings company is now an input to what a named model’s answer displays beside its citations; see OpenAI.
• The government customer. The Air Force contract above, the Cyber National Mission Force and Joint Research Centre work the Committee reports, and, in the other direction, the Committee’s demand letters of June and October 2024.
• The target. The FTC’s civil investigative demand, the Omnicom order and NewsGuard’s suit (below), with the Texas consent decree of 8 April 2026 barring the State Department from funding media-credibility rating tools marketed for suppressing protected speech; the decree does not name NewsGuard (see Global Engagement Center).

Actions & Leadership Choices

The PR register is “restoring trust in news.” The deeds describe a privately-owned scoring seat whose output is wired into three chokepoints at once — advertiser spend, AI trust signals, and a government counter-messaging contract.

• The score is a money gate. Through BrandGuard, NewsGuard licenses its ratings to advertisers and ad-tech as brand-safety data; a low reliability score is thus an input that can steer ad dollars away from a rated outlet — the same defund-by-rating mechanism documented across the flagging-machine apparatus, here run as a commercial product rather than a grant-funded index. (Cross-reference the Graphika and DFRLab files for the government-funded variants of the ratings-to-defunding pipeline, and the Stanford Internet Observatory file for the academic-flagging variant.)
• The score is an AI trust signal. NewsGuard licenses its ratings and its AI Tracking Center to model developers; the center’s count of “Unreliable AI-Generated News” sites grew from 49 in May 2023 to 3,749 across 16 languages by the current tracker. The rating a human analyst assigns to an outlet becomes a weight inside systems that decide which sources a model trusts — the ratings seat reaching into model training and grounding.

• Its account of its work for WHO. Its own case study says that “Throughout the pandemic, NewsGuard has provided the WHO with reports highlighting the top COVID and vaccine myths circulating across the major social platforms, the sources responsible for advancing those claims, and the groups on which those claims thrive,” and that, using these insights, “the WHO alerts the social media platforms of the hoaxes detected by NewsGuard, prompting them to take down posts or flag them as unreliable.” It prints a quotation from Andy Pattison, Head of the World Health Organization’s Digital Channels Team: “NewsGuard has been instrumental in helping WHO and partners keep people safe by identifying and combating online COVID-19 and vaccine misinformation” (NewsGuard, WHO case study).

• The score drew a congressional demand and a First Amendment suit. In October 2024 House Oversight Committee Chairman James Comer demanded NewsGuard disclose all government contracts amid censorship concerns. In 2023 the outlet Consortium News sued the United States and NewsGuard (second amended complaint, October 2023), alleging that under the Misinformation Fingerprints government contract the two “acted to… publicly label, target and stigmatize news organizations as disfavored” in violation of the First Amendment; the suit notes NewsGuard flagged six of the tens of thousands of articles Consortium News had published. On 26 March 2025 U.S. District Judge Katherine Polk Failla dismissed the suit with prejudice, holding the ratings protected opinion and finding no government nexus for the First Amendment claims; Consortium News did not appeal and settled NewsGuard’s legal fees for $65,000. NewsGuard denies government influence over its ratings.

Leadership choices. The company is led by its co-CEOs, Brill and Crovitz, who have, publicly, both taken the government contract and publicly defended it as narrow and foreign-facing. There is no recorded instance of NewsGuard declining the ratings-as-brand-safety business on conflict grounds; the model — a private scorer paid by advertisers and, on a single-digit share, by the Pentagon, whose score gates the revenue of the outlets it rates — is the product, operated in the open.

The suit, the appeal and the chatbot, current. On 6 February 2026 NewsGuard sued the Federal Trade Commission and Chairman Andrew Ferguson in the District of Columbia (No. 1:26-cv-00353, Judge Dabney Friedrich), represented by the Foundation for Individual Rights and Expression, seeking to quash the civil investigative demand the FTC issued in May 2025 and to block enforcement of the Omnicom merger condition. The FTC withdrew the demand by letter, announced in a filing of 17 April 2026. On 22 April 2026 the court denied the preliminary-injunction motion without prejudice; NewsGuard noticed an interlocutory appeal the same day (D.C. Cir. No. 26-5138), and on 24 June 2026 the district court stayed proceedings pending that appeal (CourtListener docket; NewsGuard, Why We’ve Just Sued the Government). In the same months the company launched a news chatbot on 23 June 2026 and, on 21 September 2026, an edition for Europe in French, German and Italian; on 12 March 2026 it launched an “AI Content Farm” detection data stream (press releases listed on NewsGuard’s press page).

CONDUCT: RATINGS-AS-CHOKEPOINT — PRIVATE SCORE, PUBLIC-CONTRACT REACH. NewsGuard does lawful, disclosed rating work by trained analysts — and it is structurally a private credit bureau for news whose single number flows into advertiser spend, AI trust weights, and a Defense Department counter-messaging contract at once. The ratings may be sound; the shape is a chokepoint one company controls.

Alliance Map

These run at the same time. One company is a for-profit ratings vendor owned by its two journalist founders and a short list of outside investors that includes a foundation and an AI-safety fund, holds a Defense contract and reported work for two other government customers, has a registered lobbyist working on platform liability, has its rating displayed inside the two largest chatbots, and is in litigation with the federal agency that investigated it and whose merger order, as NewsGuard describes it, bars a major ad-agency holding company from its feed. NewsGuard files each of these itself: its investors page, its press releases, its lobbying registrations and its complaint. The overlap is the finding.

Reach Assessment

Institutional: High. The rating is a product, not a report: licensed to advertisers and platforms, sold to AI developers, shown as an overlay in two chatbots and, since June 2026, behind a chatbot of its own. The federal government has been both a customer and an adversary, and NewsGuard says the agency reached its advertiser feed through a merger-order condition.

Memetic: High. The 0 to 100 score and the “Nutrition Label” are the company’s own shorthand for a source’s reliability, and the AI False Claims Monitor publishes a recurring percentage for how often leading chatbots repeat false claims.

Civilizational: Moderate. NewsGuard does not train or deploy a frontier model and writes no refusal policy. It rates the sources that models cite, sells the developers a feed of what to avoid, and builds a chatbot that cites only what it rates. The hazard is the position, not the people: one private company’s judgment of reliability becoming an input to many systems at once. The score holds at 76 (8/6/7), above CCDH (8/6/6, 72) and below GDI (8/6/6, 77): the entrenchment of 7 rests on an installed licensing base and now product integrations, the funding insulation of 6 on a mix of subscriptions, licensing and a government contract disclosed as a small share, and the reach of 8 on the number’s presence inside advertising and AI answer displays.

Sources: About — NewsGuard; NewsGuard AI Tracking Center; NewsGuard wins Pentagon/State Department contest — NewsGuard press release; USASpending.gov — DOD contract to NewsGuard Technologies (FA864921P1569); Matt Taibbi, “NewsGuard Case Highlights the Pentagon’s Censorship End-Around,” Racket News; House Oversight Committee — Comer Demands NewsGuard Disclose All Government Contracts; US Government, NewsGuard Sued by Consortium News — Consortium News, 23 Oct 2023.

Further sources: Our Investors — NewsGuard; Our Story — NewsGuard; Free Speech and Censors — NewsGuard; Why We’ve Just Sued the Government — NewsGuard, 6 February 2026; NewsGuard’s Reliability Ratings Now Appear in ChatGPT and Gemini Responses — NewsGuard, 12 May 2026; NewsGuard AI; FAILSafe for AI — NewsGuard; NewsGuard press releases; NewsGuard Technologies v. FTC, D.D.C. docket — CourtListener; NewsGuard Technologies v. FTC, D.C. Cir. docket — CourtListener; Consortium for Independent News v. NewsGuard Technologies, docket — CourtListener; Comer Demands NewsGuard Disclose All Government Contracts, 25 October 2024 — House Oversight Committee; Lobbying Disclosure Act filings for NewsGuard Technologies — lda.gov; USAspending — contract FA864921P1569; FARA registrants — efile.fara.gov.