The Justice Department’s crackdown on employers who favor visa sponsored hires over U.S. workers has been running quietly through mid size manufacturers, agricultural contractors, and back office operations since 2025. On August 4, 2026, it landed on an employer that builds the hiring software the rest of the industry is racing to adopt: OpenAI, alongside its analytics subsidiary Statsig, agreed to pay $3.2 million to resolve a federal finding that the two companies shut U.S. applicants out of well paid, green card track roles.

What DOJ Found at OpenAI and Statsig

The Civil Rights Division’s investigation focused on how the companies recruited for positions tied to the Permanent Labor Certification process, the step an employer must complete before sponsoring a foreign worker for a green card. Federal law requires that recruitment for those roles be conducted in good faith, on terms comparable to how the employer normally hires. Investigators found OpenAI did not advertise some PERM linked openings on its own careers site even though it posted other roles there, required U.S. applicants for those same positions to apply by mail while everyone else could apply online, and in some cases ran the job ads as late night radio spots, a channel unlikely to reach the audience an employer would use if it actually wanted U.S. applicants to see the listing.

Fewer than ten positions were involved, but the settlement is not small: $1.2 million in civil penalties plus a $2 million fund for back pay to workers who were screened out. “It is illegal to discriminate against U.S. workers by preferring temporary visa holders for jobs,” said Assistant Attorney General Harmeet K. Dhillon of the Civil Rights Division. OpenAI and Statsig denied the allegations and the underlying findings as part of the settlement, which is standard in these agreements, but both companies agreed to three years of DOJ monitoring, a rewritten recruitment policy, staff training on immigration anti-discrimination law, and reports to the department every six months.

Advertisement

HRTech Your brand belongs here. Reach the decision-makers who read HRTech every day. Premium placements across the site and newsletter. Advertise with us

A Pattern, Not an Isolated Case

The mechanism matters more than the dollar figure. This is the thirteenth settlement under the Protecting U.S. Workers Initiative since DOJ relaunched it in 2025, a program that originated in Trump’s first term in 2017 and went dormant under the Biden administration. Facebook paid $4.75 million in penalties plus $9.5 million in back pay in 2021 for nearly identical conduct, mail only applications and a refusal to seriously consider U.S. candidates for PERM roles. Apple paid $6.75 million plus $18.25 million in back pay in 2023 for paper applications and unadvertised postings. Cloudera is currently in active litigation over a PERM intake email DOJ alleges never worked. A Mississippi agricultural staffing firm, H2A Complete II, was the program’s second settlement after the 2025 relaunch. OpenAI and Statsig’s case is now the marquee example: an AI company, hiring at speed to keep up with a talent war it helped create, and one whose own products are already reshaping what HR departments do all day, tripped over the same compliance gap that has caught employers across industries for five years running.

What has changed is the target list. DOJ’s enforcement guidance now explicitly flags a newer tactic alongside the familiar mail only applications and dead end intake systems: AI generated job advertisements that are effectively visa restricted, meaning the listing’s language or placement functionally excludes U.S. applicants even without saying so. For a publication covering how AI hiring tools are becoming employers’ biggest litigation exposure, that detail is the real story inside this settlement. The compliance risk is no longer just how a human recruiter handles a PERM req. It now includes how the ATS, the job ad generator, and the sourcing algorithm built into modern recruiting stacks handle it on autopilot.

The Compliance Requirements Baked Into the Settlement

Beyond the payment, OpenAI and Statsig must post PERM linked roles publicly and accept electronic applications on the same terms as every other opening, train recruiting staff annually on the Immigration and Nationality Act’s anti-discrimination provisions, and submit to DOJ oversight for three years. That template, near identical to what Meta and Apple accepted, is becoming the de facto compliance checklist the department expects every employer running PERM sponsorship to already be following, whether or not they have been investigated yet.

Newsletter

Get the week's best tech coverage.

Free. Read by thousands of HR, tech, and business leaders.

What It Means for the HR Leader

Any organization sponsoring green cards, not just AI labs, should treat this as a signal to self-audit before DOJ does it for them. The specific practices DOJ has now flagged across five years of settlements are consistent enough to build a checklist from: no separate application path for PERM roles, no default to paper or mail when electronic options exist elsewhere, no dedicated intake channel that silently drops applications, and no PERM listing withheld from the same external job boards used for every other opening. Recruiting and immigration teams should run every PERM requisition through identical screening and identical channels as standard hiring, and document that process, because the interactive record is what regulators and courts now ask to see first.

The AI specific piece deserves its own line item. If a company’s applicant tracking system or an AI tool auto drafts and places job ads, someone in HR or legal needs to verify those tools are not quietly narrowing who sees a PERM linked posting, whether through keyword choices, placement timing, or channel selection the software picked on its own. That is a new kind of audit, one that sits at the intersection of recruiting operations and AI governance, and few HR teams have assigned clear ownership of it yet.

With thirteen settlements and counting, and enforcement now explicitly naming AI generated recruitment content as a risk area, the employers most exposed are the ones scaling hiring fastest around visa sponsored talent, which describes much of the AI sector right now. The companies writing the recruiting software the rest of HR runs on just proved they are not exempt from the rules governing how it gets used.

Source: Department of Justice