ANTI-DEFAMATION LEAGUE (ADL)

Institutional Archetype

THE ADVERTISER LEVER — This profile scores one mechanism and nothing else: the position ADL occupies in the trust-and-safety apparatus as a private organization that (a) organized advertisers to withhold spending from a platform until it changed its content-moderation conduct, and (b) produces the research, indices, and platform-facing reports that name what counts as “hate” online. It does not score ADL’s cause, its advocacy, or its character. The mechanism is structural: a non-platform body holds a lever over platform advertising revenue and supplies the taxonomy platforms are pressed to enforce. Whether that lever is pulled for good ends or bad ones is outside this file. That the lever exists, and that a private organization holds it, is the finding.

Mandate & Origin (T&S mechanism only)

The trust-and-safety mechanism runs through ADL’s Center for Technology and Society (CTS), which describes itself in its own words as “a research-driven advocacy center that works to end the proliferation of antisemitism and all forms of hate and harassment online.” The CTS output relevant here is not advocacy in general but the platform-facing instruments: research reports examining named platforms’ moderation conduct (its own recent examples include reports on Meta’s policy changes and on Cloudflare), and the Online Hate Index, ADL’s project to detect and measure hateful content at scale.

Funding & Backers

Not scored here. ADL’s general funding and governance are outside the T&S-mechanism scope of this file, and no funding claim is load-bearing to the trusted-flagger / advertiser-pressure position. The one funding fact that is on-mechanism is that the Stop Hate for Profit campaign was a coalition effort — ADL holds the campaign’s trademark, and the coalition’s named partners include Color of Change, the NAACP, Common Sense Media, Free Press, LULAC, the National Hispanic Media Coalition, Mozilla, and Sleeping Giants — i.e., the lever was pulled collectively, not by ADL alone.

The filing, current. ADL’s Form 990 for the year ended 30 June 2025 reports total revenue of $135,300,377 against $162,931,957 the year before, expenses of $128,163,819, net assets of $50,454,946 and 623 employees. Contributions and grants were $135,485,438: $17,028,435 from fundraising events, $12,821,625 from related organizations and $105,635,378 from other gifts. The government-grants line is blank, and program-service revenue was $427,368. The governing body is 21 voting members, 20 of them independent, and CEO and National Director Jonathan Greenblatt is listed at $1,127,131 of reportable compensation from ADL and $198,905 from related organizations (ProPublica, Form 990, year ended June 2025). Schedule R lists four related tax-exempt organizations, the Anti-Defamation League Foundation, the ADLF Common Fund, JLENS and Anti-Defamation League - Israel in Jerusalem, and one disregarded entity, Holocaust Organization for Promoting Education LLC (Schedule R).

The federal ledger, from the government’s side. USAspending, queried for recipients named “Anti-Defamation League” across all federal awards, returns five small contracts and no grants: $9,000 (2014) and $7,000 (2017) for Coast Guard Academy cadet diversity peer-education training, $12,000 (2018) for cadet professional-development training, a $24,800 Homeland Security training contract in 2006 held by the Anti-Defamation League Foundation, and a $0 Justice Department community-training award in 2020, $52,800 in all (USAspending HSCG3914PPSS174; 70Z03918PPSS31700; HSHQDC06P00180). The money on the record is a rounding error against the budget. The government tie that matters runs through lobbying and through the law-enforcement channel, both below.

Institutional Voice & Intent

The register is civil-rights institutional: first-person plural, historical, and sure of its own credibility. ADL’s own site says, “ADL is the leading anti-hate organization in the world” (ADL, Who We Are). Its Form 990 states the mission as “ADL WAS FORMED IN 1913 TO STOP THE DEFAMATION OF THE JEWISH PEOPLE AND TO SECURE JUSTICE AND FAIR TREATMENT TO ALL.” Greenblatt gave the House Energy and Commerce subcommittee the technology arm’s pitch on 9 December 2021: “In 2017 ADL launched the Center for Technology and Society to double down on our efforts to fight online hate,” a unit “staffed not by longtime non-profit professionals, but by software engineers, product managers, data scientists, and computer experts, all hired from industry.” The policy ask came in the same sitting: “Section 230 must be changed to force the companies to play by the same rules that every other media company on the landscape operates by today” (House hearing record, 9 Dec 2021). When the FBI Director made an announcement on 1 October 2025 regarding the Bureau’s relationship with ADL, ADL answered in three sentences: “ADL has deep respect for the Federal Bureau of Investigation and law enforcement officers at all levels,” and “In light of an unprecedented surge of antisemitism, we remain more committed than ever to our core purpose to protect the Jewish people” (ADL statement, 1 Oct 2025).

Stated alignment: Stop “the defamation of the Jewish people” and “secure justice and fair treatment to all.” Observed alignment: A lawful advocacy nonprofit that runs a platform-facing research arm, holds a trademarked advertiser campaign, lobbies on platform-liability and transparency bills, and now benchmarks AI models by name. Gap assessment: The cause is outside this file. The file scores the lever, and the record shows the lever is used in the open, with the budget, the lobbying spend and the research all published.

Stated intent: “A research-driven advocacy center that works to end the proliferation of antisemitism and all forms of hate and harassment online,” in partnership “with industry, civil society, government and targeted communities” (ADL, Center for Technology and Society).

Observed intent: Hold platforms to account for hate on their services, and be the party that measures it. The Center publishes the audits, the Stop Hate for Profit campaign supplied the pressure, and the lobbying filings name the bills. Interest, as documented: the Section 230 change Greenblatt asked for in 2021 and the platform bills ADL’s filings list (the STOP Hate Act of 2025 and the Platform Accountability and Transparency Act of 2025, by the titles the filings give) would put duties on platforms, and the Center already produces the platform audits that such duties would formalize. ADL’s own best case, in Greenblatt’s words: it has been “tracking online hate since the days of dial-up,” and “this isn’t a resource problem, it is a responsibility problem.” The same Section 230 reform appears in the lobbying registrations of two other subjects of this site, CCDH and NewsGuard, each through an outside lobbying firm. The convergence is the finding, and no coordination among them is asserted.

Gap: ADL’s defense is its record: a century of tracking, a disclosed budget and a public paper trail. The record cannot settle whether any platform changed a rule because ADL asked, and it cannot settle whether the platform bills would work as intended. For the lever, it never needs to.

• Social accounts and feeds. X @ADL, Facebook, Instagram, LinkedIn and YouTube, all linked from adl.org; feed adl.org/rss.xml (HTTP 200, RSS; not advertised in the page head).

Position in the Apparatus

ADL sits in five seats of the apparatus at once, and only the first two were in the original file.

• The advertiser lever. Stop Hate for Profit, trademarked by ADL, asked advertisers to pause Facebook and Instagram spending in July 2020; see the section above and Meta. The campaign page still carries the line “Stop Hate For Profit is a trademark of the Anti-Defamation League” (Stop Hate for Profit).
• The taxonomy and the ratings. The Center’s current output names platforms and products: reports on Wikipedia’s handling of biased edits, on AI chatbots and Persian speakers, on Cloudflare, on open-source AI models’ guardrails and on AI video generators (ADL, Center for Technology and Society).
• The AI benchmark. ADL’s report on anti-Jewish and anti-Israel bias in large language models queried GPT (OpenAI), Claude (Anthropic), Gemini (Google) and Llama (Meta): “Each LLM was queried 8,600 times for a total of 34,400 responses.” It reports that all four showed bias, that “Llama demonstrates pronounced anti-Jewish and anti-Israel biases” as “the lowest scoring model for both bias and for reliability,” and that “GPT and Claude show the most anti-Israel bias of any of the models tested” (ADL, Generating Hate). A private advocacy organization has written the scorecard on which named labs (OpenAI, Anthropic, Meta) are graded for a category of bias. The instrument is ADL’s own agreement-scale design, and its findings are ADL’s.
• The law-enforcement channel. Greenblatt told the House subcommittee that “in the past 11 months, we have provided the FBI with more than 1,000 actionable tips” (hearing record, above). On 1 October 2025 the FBI Director made an announcement regarding that relationship, per ADL’s statement above.
• The legislative seat. ADL files its own lobbying disclosures (below) and posts its submissions to federal bodies, among them its comments on the Office of Management and Budget’s federal AI governance request, on the NTIA’s AI accountability request, on the President’s Council of Advisors’ generative AI request and a written statement to the Senate AI Insight Forum on elections and democracy (ADL, letters, testimony and comments).

Actions & Leadership Choices (T&S mechanism only)

The mechanism, read on the record, has two moving parts.

The advertiser lever. In June 2020, ADL and coalition partners launched Stop Hate for Profit, which — in the campaign’s own words — asked advertisers to “pause their spending on Facebook and Instagram ads for the month of July 2020” to “convince social media companies to finally put people over profit.” The campaign’s stated target was, verbatim, “hate, bigotry, racism, antisemitism and disinformation.” Hundreds of advertisers paused; by the campaign’s own account the pressure yielded concessions from the platform, including a civil-rights executive role and the release of a delayed civil-rights audit. The mechanism is the point: a coalition of private advocacy organizations demonstrated it could move a platform’s moderation conduct by moving its advertising revenue.

The flagging/ratings instrument. Separate from the boycott, ADL supplies the classificatory layer: reports grading or naming platform conduct on hate, and the Online Hate Index as a scaled detection instrument. This is the trusted-flagger half of the position — the organization does not run any platform’s moderation queue, but it produces the research and the taxonomy that platforms are publicly pressed to adopt and enforce.

On GARM participation. [SOURCE NEEDED] — a primary source establishing ADL’s formal participation in the Global Alliance for Responsible Media was not obtainable this session (WebSearch budget exhausted; no ADL or WFA primary page confirmed). No GARM membership is asserted. The advertiser-pressure finding above rests on Stop Hate for Profit, which is fully sourced.

The platform bills, the AI desk and the courtroom, current. ADL’s own lobbying disclosures list, under Science/Technology, work “related to reducing hate in digital social spaces and artificial intelligence,” naming the STOP Hate Act of 2025 and the Platform Accountability and Transparency Act of 2025 (LDA filing, Q2 2026). The Center’s AI work moved from reports on platforms to benchmarks of models, above. In the courts ADL is more often the defendant than the plaintiff: Louis Farrakhan and the Nation of Islam sued ADL, the Simon Wiesenthal Center, Abraham Cooper and Greenblatt in the Southern District of New York (No. 1:23-cv-09110), and on 5 April 2024 the court granted the motions to dismiss and dismissed the second amended complaint in its entirety; the plaintiffs appealed to the Second Circuit (No. 24-1237) (CourtListener, SDNY docket; Second Circuit docket).

The outside lobbyist. Beside its own filings, the Senate lobbying database lists Ballard Partners as a registrant for the ADL: registered 2 January 2025 under the issue line “Strategies and development of practices and policies to counter the rise of antisemitism in the U.S.,” reporting $10,000 for Q4 2024 and $40,000 in each of Q1 to Q3 2025, lobbying the Justice and Education Departments, and terminated on a form dated 15 October 2025 (LDA, Ballard Partners registration for ADL). See Ballard Partners.

CONDUCT: FLAGGER — ADVERTISER-PRESSURE POSITION. A private advocacy organization that (1) trademarked and co-led an advertiser campaign shown to move a platform’s moderation conduct by threatening its ad revenue, and (2) produces the platform-facing hate research and detection index that feed the classification layer of trust-and-safety. The lever and the taxonomy are the finding; ADL’s cause, motive, and character are not scored, and the nation-state-not-ethnic rule is held absolutely — nothing here turns on the substance of what ADL advocates, only on the structural position it occupies. Compare the ratings-and-flagging position mapped at NewsGuard.

Alliance Map

These run at the same time. One organization holds an advertiser trademark, files seven-figure lobbying disclosures, reports no government grants on its Form 990, benchmarks four labs’ models for bias, and supplies tips to the FBI, a relationship the FBI’s Director has since addressed publicly. ADL files each of these itself: the budget, the lobbying and the statement of mission are all its own documents. The overlap is the finding.

Reach Assessment

Institutional: High. The lever demonstrated in 2020 was an advertiser pause that the campaign says won concessions; the audit seat now reaches AI models; the lobbying runs at $1.49 million for 2025 by the filer’s own quarterly figures; and the FBI relationship was real enough for the Director to make an announcement about it. Its independence is not in doubt on the money: $135 million of contributions, no government grants, 20 of 21 voting directors independent.

Memetic: High. ADL’s annual online-harassment survey and its extremism and platform reports are the numbers that others quote. The LLM benchmark added a quotable ranking of named models.

Civilizational: Moderate. ADL does not build or deploy a model and writes no refusal policy. It grades them, in one category, with its own instrument, and lobbies on the rules that would bind the platforms that host them. The hazard is the position, not the people: a private body that holds the advertiser lever, the platform taxonomy and a model scorecard at once. The score holds at 66 (7/6/7) below NewsGuard (8/6/7, 76) and CCDH (8/6/6, 72): ADL’s reach is wide across instruments, but the advertiser lever was a one-month campaign and the audits and scorecard bind no one.

Sources: Stop Hate for Profit — campaign site; Center for Technology and Society — Anti-Defamation League; Online Hate and Harassment: The American Experience — ADL; Online Hate Index — ADL (live; bot-blocked to automated fetch, 200 snapshot at Internet Archive, 12 Jul 2024).

Further sources: Anti-Defamation League, Form 990 for the year ended June 2025 — ProPublica; Schedule R — ProPublica; Who We Are — ADL; ADL Statement on the FBI Director’s 10/1 Announcement; Generating Hate: Anti-Jewish and Anti-Israel Bias in Leading Large Language Models — ADL; ADL letters, testimony and comments to Congress and federal agencies; Stop Hate for Profit; House Energy and Commerce Subcommittee hearing, 9 December 2021 — congress.gov; Lobbying Disclosure Act filings, Anti-Defamation League — lda.gov; USAspending — recipient search, Anti-Defamation League; Farrakhan v. Anti-Defamation League — CourtListener; FARA registrants — efile.fara.gov.