The PERM Question Behind the OpenAI–Statsig Settlement
The subject of this series is a 2026 settlement between the Justice Department’s Immigrant and Employee Rights Section (IER) and two companies, OpenAI OpCo, LLC and Statsig, Inc., concerning recruitment connected to the permanent labor certification program, commonly called PERM. The specific question is narrower—and more consequential—than whether a company used paper: when IER says U.S. workers had to apply for PERM-related jobs by mail while other applicants could use the companies’ external career websites, did that administrative detour change who had a fair chance to enter the recruiting pool? This series gives readers the background to separate that candidate-facing issue from an employer’s own PERM filing, read the settlement at the strength its evidence supports, and understand why equal access and searchable records became central remedies.
The case is about a route into consideration
The settlement’s background says IER opened several investigations of OpenAI and one investigation of Statsig. IER says those investigations found reasonable cause to believe that OpenAI engaged in a pattern or practice of citizenship-status discrimination in five PERM-related advertisements, beginning no later than June 2023 and continuing until at least April 2025. It says Statsig engaged in one PERM-related recruitment in July 2025. The described conduct was similar in the two accounts: U.S. workers could not apply in the same manner as people applying to the companies’ other jobs. Instead, the companies required U.S. workers to submit applications by mail.
That description makes the application route the important fact. A requirement to mail a document is not automatically unlawful, and the sources do not establish that every paper process creates discrimination. The concern arises from the comparison. If one group is directed to an external website while another group is sent to a different or more burdensome channel for the same broad recruiting purpose, the difference can affect whether people apply and whether their applications remain in the ordinary recruiting system. In the settlement’s account, the additional burden was not required by the PERM process and did not occur in non-PERM recruitment.
The distinction between an agency’s account and a final judgment matters here. The agreement says IER found “reasonable cause to believe” the described statutory violations occurred. OpenAI denies and continues to deny IER’s allegations and findings. The Parties also agree that the settlement is not an admission of a violation, wrongdoing, or liability. The document therefore gives us two things at once: a first-party account of what IER investigated and the controls the companies agreed to adopt, and an explicit warning not to describe the negotiated resolution as a court’s final determination of the merits.
That is not a reason to treat the agreement as empty. A settlement can show which failure mode a regulator considered serious enough to investigate, which facts the parties used to define the problem, and which operational changes they considered necessary to resolve it. It can also show what the record does not tell us. The two sources do not provide the full text of the five OpenAI advertisements or the Statsig recruitment, the number of people who encountered the mail requirement, the final policies, or individual back-pay outcomes. The analysis must remain specific without pretending to contain a complete case file.
Why PERM makes access a substantive question
The Department of Labor’s PERM guidance supplies the institutional background. A permanent labor certification allows an employer to hire a foreign worker to work permanently in the United States. In most cases, before the employer submits an immigration petition to U.S. Citizenship and Immigration Services, the employer must obtain a certified labor certification application from DOL’s Employment and Training Administration. DOL says the certification must establish that there are not sufficient U.S. workers able, willing, qualified, and available to accept the job in the area of intended employment, and that employing the foreign worker will not adversely affect the wages and working conditions of similarly employed U.S. workers.
Those two propositions make the U.S.-worker pool part of the program’s substance. Recruitment is not only a company’s internal search for a convenient set of candidates. It is part of the factual setting in which the availability of U.S. workers is assessed. If a recruiting design makes it materially harder for some U.S. workers to enter that pool, the design can matter to the reliability of the underlying labor-market question. That is a relationship supported by the program’s stated purpose; it is not, by itself, a conclusion that a particular employer violated the law.
DOL’s wording also tells us why access comes before selection. A person who cannot or does not reach the application stage cannot be evaluated for willingness, qualifications, or availability in the same way as someone who enters the employer’s normal system. The public guidance does not explain how each of those words is measured in every occupation, and it does not tell us how many applicants a particular employer should receive. It does establish the direction of the inquiry: the program is concerned with whether suitable U.S. workers are present in the relevant labor market, not merely with whether an employer can document a completed filing.
The second part of DOL’s statement broadens the stakes. The certification is also tied to wages and working conditions for similarly employed U.S. workers. That means recruitment cannot be understood only as a private contest for one job. The way an employer identifies and considers U.S. workers sits alongside the program’s concern about the effect of the eventual foreign-worker employment on comparable workers. The sources do not let us calculate that effect here, but they make clear why a barrier at the entrance to recruitment deserves attention before anyone debates the final hiring decision.
This is why the word “friction” should be used carefully. A website that is slow, a form that asks for an extra field, or a paper route that is available to everyone may raise practical concerns, but those facts alone do not establish the issue described in the settlement. The narrower concern is a PERM-specific route that the employer does not impose in ordinary recruitment and that IER says was imposed on U.S. workers. The burden becomes relevant because it is connected to who can reach consideration, not because convenience is always a legal requirement.
The settlement’s account turns that general program logic into a concrete comparison. It does not say only that the companies’ process was inconvenient. It says U.S. workers were not allowed to submit applications in the same manner as other job applicants, and that the added mail requirement was not needed for PERM recruitment. The comparison therefore supplies a possible mechanism between policy and outcome: a different route can change who takes the step of applying, while a route outside the ordinary recruiting system can make those applications harder to see. That is an explanation of why the practice matters, not proof from these sources that a specific person was excluded or that a particular decision-maker acted with a particular intent.
The practical judgment this case supports is therefore a sequence of questions. Who had to do what to apply? What route did the employer normally use for comparable jobs? Was the extra step required by the government process, or was it an internal choice? After an application arrived, could the recruiting team find it and evaluate it under the same job-related standards? These questions connect a candidate’s first interaction with the company to the later evidence an employer would need to show that the process operated fairly. They also keep the analysis anchored to observable differences rather than asking the reader to infer a motive that the settlement does not establish.
The word “mail” hides two different decisions
The DOL page contains a detail that can easily send the discussion in the wrong direction. At the final submission stage, an employer can submit a PERM application online after registering with Login.gov and creating a FLAG account. DOL also says employers may mail in their applications. The agency presents online submission as a way to streamline processing, improve access for authorized users, assist OFLC, and provide an immediate copy of a certified labor certification. In this context, “mail” describes an employer sending an application and supporting documentation to the government.
The settlement discusses something else. It says U.S. workers applying for PERM-related positions had to submit job applications by mail instead of applying electronically through the companies’ external job websites. The actors, destination, and stage are different. One route is an employer’s filing with OFLC; the other is a worker’s attempt to enter the employer’s recruiting process. The fact that DOL allows an employer to mail its own PERM application does not show that DOL required a worker to mail a job application.
The relevant comparator is ordinary non-PERM recruitment. The settlement says the mail burden did not occur there. That comparison changes the question from “Is paper ever used?” to “Why did this category of position use a different route for U.S. workers?” A company may need to comply with a particular government form or agency rule. But if the government does not require the candidate-facing detour, and the employer’s ordinary recruiting system already accepts electronic applications, the detour can become an organizational choice with unequal consequences.
The distinction also explains why the settlement’s remedy is more detailed than a promise to stop discriminating. The agreed policy must give each PERM-related position a unique identifier, make it searchable on internal and external career websites, and accept electronic applications through those pages in the same manner as non-PERM recruitment. Those requirements respond to the precise failure described by IER: a job that exists for labor-certification purposes must also be findable and usable in the recruiting system where U.S. workers apply.
The question to carry forward
The opening lesson is not that every administrative burden is discrimination, or that every settlement payment proves liability. It is that a route into consideration can become a substantive compliance issue when it is tied to U.S. workers, differs from the employer’s ordinary recruiting route, and is not required by the PERM process. OpenAI and Statsig remain the named subjects; PERM’s labor-market purpose supplies the reason the route matters; IER’s settlement supplies the documented account and the remedial design; and the denials and non-admission clauses set the limits on certainty.
Before judging the recruitment practice, we need to see the question PERM itself is designed to answer and the basic sequence in which an employer defines a job, obtains a prevailing wage determination, conducts pre-filing recruitment, and submits an application. That process map will make it possible to keep the employer’s filing logistics separate from a worker’s chance to apply—and to understand why the difference matters in the OpenAI–Statsig settlement.