1. Who They Are
Scale AI, founded in 2016 by Alexandr Wang, built the data-labeling layer of the AI boom: armies of workers on its Outlier and Remotasks platforms annotating images, writing example responses, and rating chatbot outputs for OpenAI, Meta, the Pentagon and others. In June 2025 Meta invested roughly $14.3 billion for a 49 percent stake — valuing Scale above $29 billion — and Wang departed to lead Meta Superintelligence Labs. Wang is profiled in a separate executive fact sheet.
2. What the Book Documents
The business model, in the book's words: "Scale AI built its entire business on one specific capability: harvesting workflow data from contractors to train AI systems." The book quotes Wang's own 2024 statement of the motive — "the biggest blocker is actually a lack of data… we have to figure out how to produce really high quality data" — and documents how that logic, imported into Meta as the MCI program, turns workers' keystrokes into the training data for the systems that will replace them.
3. Verified Labor Record — Reported in Full
| Date | Proceeding | Outcome | Basis | Status |
|---|---|---|---|---|
| Aug 2024 – May 2025 | US Department of Labor (FLSA investigation) | No action — DROPPED | Investigated Scale (and HR partners HireArt and Upwork) over alleged unpaid wages, worker misclassification, and retaliation | CLOSED with no findings announced, May 9, 2025 — reported per the full-record rule; Scale welcomed the decision |
| Dec 2024 – 2025 | McKinney et al. v. Scale AI (San Francisco Superior Court) + related actions | $12.5 million class settlement | California workers — Generative AI Annotators, Prompt Engineers, Subject Matter Experts on Outlier/Remotasks — alleged they were misclassified as independent contractors despite Scale's control over their work, failing California's ABC test; a companion PAGA suit (Rogowicz) alleged effective pay of ~$15/hour, below California's minimum wage; a further suit by Outlier workers alleged assignment to write disturbing violent and abusive prompts without psychological safeguards | Settlement of the consolidated actions pending court approval; Scale disputed all claims and denies wrongdoing; automatic payments to class members if finally approved |
Sources: TechCrunch (DOL closure) · King & Siegel ($12.5M settlement; ABC test) · Official settlement notice summary · Inc. (probe scope; psychological-harm suit)
Context the record supports: a 2023 Washington Post investigation documented overseas Scale contributors — in Kenya, the Philippines and beyond — complaining of low pay and impossible demands; worker communities documented rate compression on Outlier from roughly $40–50 to $15–20 per hour and unpaid onboarding time (community-reported; Scale says its rates meet or surpass local living-wage standards). Scale's Washington integration deepened through the same period: Wang testified before the House Armed Services Committee (July 2023), attended the January 2025 inauguration, and Scale's former managing director Michael Kratsios was nominated to lead White House science and technology policy.
4. Honest Accounting
Per the full-record rule, both directions: the federal investigation closed without action, and the class settlement — like all settlements — is not an adjudication of wrongdoing. What stands is the structural fact the book documents: the labor that trains the models is classified out of employment protections, priced at the margin, and — in the workers' own filed allegations — exposed to the same traumatic content economy documented at Sama, while the company's valuation rose past $29 billion on their output. Developments, including final settlement approval, will be added as verified.
5. Associations Noted in the Book
Alexandr Wang (founder; now Meta's Chief AI Officer) and Meta Platforms (49% owner) are profiled separately. Associations are noted for context; every company and person is documented on its own record.
Legal Disclaimer
This document is published as journalism and commentary in the public interest, protected as free speech and freedom of the press under the First Amendment to the United States Constitution, Article 19 of the Universal Declaration of Human Rights, and Article 10 of the European Convention on Human Rights. It compiles information from public records, court filings, government announcements, parliamentary and congressional investigations, and reporting by established news organizations, together with the documented first-person accounts and analysis published in the books No Ethics in Big Tech and Ethics in Tech and Lack Thereof by Vahid Razavi.
Where a fine, judgment, or settlement is described, its status (paid, under appeal, overturned, or pending) is stated as reported by the issuing authority or by reputable press as of the verification date above. Allegations in open lawsuits and investigations are exactly that — allegations — and every party is entitled to the presumption of innocence unless and until a competent court rules otherwise. Where a company or executive has prevailed in court, denied an allegation, or had a claim dismissed, that outcome is reported here as well, because the credibility of this project rests on stating the record in full. Opinions and characterizations drawn from the books are the protected opinion of the author. Corrections supported by documentation are welcome via NoEthicsInBigTech.com/contact.
Research Credit
Deep research, source verification, and document preparation for this fact sheet were performed by Claude, the AI assistant built by Anthropic, working under the direction of author Vahid Razavi. Every fine amount, case citation, and status line above was checked against the primary or press source listed beside it on the verification date. This is a living document; new rulings, fines, and investigations will be added as they are verified.