Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsSome links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Workers who helped evaluate and improve AI systems through Scale AI’s Outlier platform alleged that they were treated as independent contractors while facing employee-like controls, unpaid work time and, in some assignments, psychologically disturbing material. The first widely reported case was filed in California in 2024; related cases later raised wage and safety claims. The McKinney wage-and-classification dispute has since entered a settlement process, but the available settlement materials do not establish a final merits finding or show that every Outlier contributor is covered.
Who the workers are—and what they do
Scale AI provides data-labeling, evaluation and human-feedback services used in AI development. Outlier is a contributor-facing platform through which some people perform that work. Tasks can include rating model responses, writing or ranking prompts, labeling data and assessing whether an AI system handles safety-sensitive requests appropriately. “AI training” is a broad shorthand: the work may involve annotation, model evaluation, reinforcement-learning feedback or red-teaming rather than building a model from scratch.
The lawsuits concern particular workers, companies and projects—not every Scale AI customer or every contributor on Outlier. The later federal complaint named Scale AI, Smart Ecosystem and Outlier AI as defendants. Workers’ actual arrangements, task types and legal rights can vary.
What the first lawsuit alleged
Former contributor Steve McKinney brought a California action in October 2024; Computerworld reported on it on December 12. The complaint alleged that workers were classified as independent contractors despite controls over how work was assigned and performed. According to the reporting, workers said algorithmic task assignment, time limits and quality decisions could affect whether they were paid, and that compensation could be reduced or denied when a task took longer than its designated allowance.
#1 Best Overall
The allegations also focused on time around the task itself: reading instructions, seeking clarification and attending required training webinars. The complaint reportedly said workers faced restrictions on breaks and outside research and could lose access to the platform after raising concerns about pay or conditions. Some assignments were described as involving subjects such as suicidal ideation and violence. These are allegations, not findings that Scale AI violated the law. Computerworld’s report summarizes the original claims and the company’s response.
Why classification can affect a worker’s finances
Being called a contractor in a platform’s terms does not, by itself, settle a worker’s legal status. Employees generally have wage-and-hour protections, including minimum-wage and overtime rules where applicable, as well as other statutory rights that contractors may not receive. Independent contractors typically bear more responsibility for taxes and business expenses, and may lack employee benefits and some employment-law remedies.
California uses an ABC test in many worker-classification disputes. In general, a hiring entity must establish that the worker is free from its control, performs work outside its usual business, and is customarily engaged in an independently established trade or business. Whether the test applies and how it comes out depend on the facts, the claims and any relevant exemptions or agreements. The test does not automatically decide these cases.
Recommended Free Tools
For a worker assessing a platform job, the practical questions go beyond the contract label: Who sets the rules and time limits? Can a worker turn down tasks without penalty? Is the work central to the company’s business? Is there a genuinely independent business? Are required instruction, training and administrative tasks paid? Those facts can matter alongside state law, arbitration terms and the location where the work was done.
Rank #2
The pay question: count the work around the task
A displayed hourly rate or maximum rate is not necessarily a guarantee of that amount for every hour spent working. A worker’s effective rate depends on what time is paid, how tasks are priced, whether rejected work is compensated, and how much work is available.
A separate California representative and PAGA complaint filed by Amber Rogowicz on January 3, 2025, alleged that required instruction and training time was not fully paid. The complaint reportedly described a typical workday of about 10 hours for which roughly five hours were compensated, producing an alleged effective rate of about $15 an hour. Those figures are the plaintiff’s allegations and calculation, not an independently established wage finding. The complaint sets out the claims; a case listing identifies the filing.
For gig and platform workers generally, useful records include task histories, start and stop times, instructions, training notices, payment statements, rejected-task explanations, support messages and appeal outcomes. Keeping records does not determine whether a worker is legally an employee, but it can help document the gap between time spent and time compensated.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →A second case raised psychological-risk claims
On January 17, 2025, six former contributors filed a separate federal complaint in the Northern District of California, case 3:25-cv-00620. Angela Schuster, Anna Pendleton, Howard Quattlebaum, Xavier Retana, Latoya Howard and McKinney alleged that work on AI-safety projects exposed contributors to psychologically harmful material, including prompts involving severe violence, suicide and sexual abuse, without adequate safeguards or medical monitoring. The complaint asserts negligence theories and a claim under California’s Unfair Competition Law. Filing a complaint does not prove its claims. The filed complaint describes the allegations, while the docket summary identifies the case and parties.
Safety testing can require people to examine adversarial or harmful prompts to see how systems respond. The labor question is how that work is designed: whether people get clear advance notice, meaningful opt-out choices, workload limits, screening and psychological support—and whether declining sensitive work jeopardizes access to other paid tasks. Scale AI’s response, as reported by Computerworld, was that sensitive projects came with advance notice and opt-out options, and that contributors had access to health and wellness programs. The plaintiffs’ claims and the company’s description of its safeguards are distinct accounts; the available sources do not establish how those measures worked for every contributor.
What happened to the cases
- October 29, 2024: The McKinney-related California action was filed, according to a case summary.
- December 12, 2024: Computerworld reported on the allegations concerning working conditions and classification.
- January 3, 2025: Rogowicz filed a California representative/PAGA complaint.
- January 17, 2025: The six-plaintiff federal psychological-harm complaint was filed.
- February 19, 2025: The federal docket summary reflected an order addressing service, responsive pleadings and briefing. That summary says it was last retrieved on that date, so it is not a reliable account of the federal case’s current status.
- April 2026: A California labor-case listing reports that the McKinney-related matters settled.
- August 18, 2026: The official settlement FAQ described administration in progress, with September 3, 2026 deadlines for exclusions, challenges and objections.
The timeline draws on the labor-case summary, the Rogowicz case listing, the federal docket summary and the official settlement FAQ. The federal docket information available here is incomplete; it does not support saying that the psychological-harm case is over.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What the settlement does—and does not—mean
A settlement process is not the same thing as a court finding that a company violated the law. A case can settle to provide compensation and avoid the cost and uncertainty of continued litigation without either side’s allegations being established at trial. Nor does a settlement for a defined group automatically cover every person who has worked for Outlier or Scale AI.
The official FAQ describes procedures for class members to challenge listed workweeks or PAGA pay periods, request exclusion or object. It also refers to final-approval and fee motions, so the FAQ supports describing the matter as in settlement administration—not as a final merits judgment. It identifies a $175,000 PAGA allocation to be divided according to PAGA pay periods; that is an allocation within the settlement materials, not the total settlement value.
Rank #4
Coverage and consequences depend on the official notice and settlement documents: the covered dates, class definition, payment formula, release language and final-approval status matter. A person who opts out may retain the ability to pursue an individual claim but would generally give up settlement benefits; a person who does not opt out may be bound by a final approved settlement and release. The precise effects depend on the governing documents and court approval.
Who may be affected
The California McKinney-related settlement applies only to people who meet its formal definition, not automatically to all current or former contributors. Other workers may be outside the covered dates or location, may have worked through a different entity or platform, or may have different contracts. State and country laws differ, as do rules for ordinary annotation, evaluation and higher-risk safety work. Arbitration or forum-selection provisions may also affect an individual’s options.
If you believe you may be included, use the official settlement FAQ and notice to check eligibility, deadlines and procedures rather than relying on a news summary. Preserve relevant pay and task records, and consider speaking with an employment lawyer or legal-aid organization for advice about your own circumstances. This article is general information, not legal advice.
Why the dispute matters beyond one platform
AI products can depend on a dispersed human workforce whose labor is largely invisible to end users. A platform can describe contributors as independent businesses while also using task allocation, quality scoring, time limits, required rules and unilateral account decisions. When work availability fluctuates and onboarding, appeals or safety exposure go uncompensated, workers may bear risks that are hard to see in a headline rate.
The lawsuits do not establish a new rule for the AI industry, and their different claims should not be collapsed into one. They do, however, surface questions relevant to data-labeling, content moderation, model evaluation and AI-safety vendors: how control works in practice, what time counts as paid work, and what protections accompany exposure to disturbing material. The settlement process may resolve claims for a defined group, while the wider classification and worker-safety questions remain fact-specific and unsettled.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
