The U.S. Department of Justice announced on August 5, 2026, that OpenAI has agreed to pay $3.2 million and accept three years of government oversight over how it hires for certain jobs tied to green cards. The DOJ had accused OpenAI and its one-time subsidiary Statsig of making it intentionally difficult for U.S. citizens to apply for those positions. Neither OpenAI nor Statsig admitted any wrongdoing.

The investigation began in August 2025 and covered five cases at OpenAI between 2023 and 2025, plus one case at Statsig. Fewer than 10 jobs were involved in total. The settlement includes $1.2 million in penalties and $2 million set aside to compensate U.S. citizens the DOJ determined were harmed.

To understand the case, it helps to know how green-card sponsorship works. When a U.S. company wants to help a foreign employee get a green card, federal law says the company must first prove it tried to find a qualified American for the job and could not. This is called the PERM process. The law behind it, the Immigration and Nationality Act, dates back to 1952.

The DOJ alleged that OpenAI and Statsig set up their PERM recruitment in a way that discouraged qualified Americans from applying. Specifically, the companies allegedly did not post the jobs on public job boards, advertised some openings on late-night radio, and required paper applications instead of letting people apply online.

According to the DOJ, the practical effect was to reduce the chances that a qualified U.S. citizen would apply and potentially get the job. If an American had filled the role, the company's plan to sponsor its existing foreign worker for a green card would have been complicated or stopped entirely.

Under the settlement, OpenAI and Statsig must write new hiring policies for these roles and get the DOJ to approve them. They must also file reports every six months showing how many people applied and how many U.S. citizens were interviewed.

The relationship between OpenAI and Statsig adds a complication. OpenAI acquired Statsig, an AI testing company, in September 2025, then sold off at least part of the business in May 2026. The settlement applies to OpenAI directly and covers alleged conduct during the period when Statsig was under OpenAI's ownership, and possibly before.

Reuters, Axios, and Fox Business all reported the $3.2 million figure on August 4 and 5, 2026. The DOJ's official press release is published on the Justice Department's Office of Public Affairs page, and the settlement is also documented on the Civil Rights Division's Settlements and Lawsuits page and its Immigrant and Employee Rights Section page.

The broader context here matters. During the Biden administration, both Facebook and Apple signed similar settlements with the DOJ over the same kind of hiring violations. Those cases, however, involved allegations that were widespread across large parts of the companies' hiring. The OpenAI case is narrower: fewer than 10 roles, a relatively small financial penalty for a company of OpenAI's size, and no admission of guilt.

The oversight terms, though, may matter more than the money. Three years of having to get DOJ approval for hiring policies and file regular reports creates a compliance burden that goes well beyond the dollar amount. For a company hiring talent from around the world, any slowdown in the green-card process can affect its ability to retain employees, especially those on H-1B work visas who need employer-sponsored green cards to stay in the U.S. long-term.

This kind of enforcement is not new. The law's protections for U.S. workers have been enforced under multiple administrations, against companies from social media platforms to hardware makers. What is different here is the target: a company at the center of the current AI boom, already under intense public and regulatory scrutiny on many fronts.

For the technology industry, the settlement is a reminder that immigration-related hiring rules operate on a separate track from the more visible debates around AI safety and antitrust. A company can be pushing the boundaries of AI research and still find itself subject to a 74-year-old law about how it advertises a handful of jobs.