OpenAI Pays $3.2M to Settle DOJ Hiring Bias Claims

OpenAI will pay $3.2 million to settle Justice Department allegations that it passed over American workers in favour of people holding temporary work visas. The DOJ's Civil Rights Division announced the settlement on Tuesday, naming OpenAI OpCo LLC and its subsidiary Statsig, the product-experimentation company OpenAI acquired last year.

The allegation is a citizenship-status discrimination claim under the Immigration and Nationality Act. The department said the two companies violated the statute by preferring workers on temporary employment visas under a federal programme that requires employers to give priority to US workers. It is the eighth settlement the department has obtained since it relaunched its Protecting U.S. Workers Initiative in 2025.

What the announcement did not say

Several things a reader would reasonably want are not in the public announcement, and it is worth being precise about that rather than filling the gaps. The department did not break the $3.2 million into civil penalties versus back pay for affected workers. It did not specify the period the conduct covers, how many positions or applicants were involved, or what remedial obligations — training, auditing, policy changes, monitoring terms — the agreement imposes. OpenAI has not issued a public response. Settlements of this type routinely carry all of those elements, but none of them are confirmed here yet.

What is confirmed is the shape of the claim, and the shape is unglamorous. This is not an AI-safety case, a copyright case, or an antitrust case. It is a paperwork-and-hiring-practice case of the kind the Civil Rights Division brings against staffing firms and IT contractors, applied for once to a company valued in the hundreds of billions. The federal programme at issue obliges an employer to test the domestic labour market before filling a role with visa-sponsored labour, and the department's position is that the test was not run the way the rules require.

The part that will travel further than the fine

For OpenAI the money is immaterial — $3.2 million is a rounding error against a company that closed the largest private venture round on record. The cost is positional. OpenAI's public argument for its own expansion, and for the data centres and power contracts that come with it, leans heavily on American jobs and American competitiveness. A federal finding that it preferred visa holders over US applicants is an awkward fact to carry into that argument, and it arrives in a season when the jobs case for AI is already being audited in public — states are clawing back tax breaks over disappointing headcounts, and the question of which jobs these systems actually displace is no longer rhetorical.

There is also a symmetry worth noting. On the same day this settlement was announced, OpenAI staff were at the White House discussing voluntary safety testing for frontier models — an arrangement with no statute behind it and no published metrics. The hiring case is the opposite kind of encounter with the federal government: an old statute, a specific rule, a named division, and a number. One of these two regimes has teeth, and it is not the one about superintelligence.