The Elegant Enterprise-Wide Solutions Settlement: Employers Answer for Job Ads, However Drafted

Stop assuming the drafter matters

If your job advertisements are drafted by software, the law still treats them as yours. That is the position the Department of Justice took in its February 2026 settlement with Elegant Enterprise-Wide Solutions, Inc. What it does not do is create a rule about artificial intelligence. The obligation did not change; the drafter did. Treat the case as a signal of enforcement posture on an existing statute, not as new law.

The binding text is 8 U.S.C. § 1324b, which has not changed. The settlement binds one company. Its terms do not mention AI, and the AI attribution rests on DOJ's press release.

The instrument and what is in force

The obligation sits in 8 U.S.C. § 1324b(a)(1). It is an unfair immigration-related employment practice for a person or entity to discriminate in hiring, in recruitment or referral for a fee, or in discharge, and (B) covers discrimination against a "protected individual" because of citizenship status³. Protected individuals are citizens and nationals, lawful permanent residents, certain temporary residents, refugees and asylees, subject to naturalisation-timing exclusions in § 1324b(a)(3)³. H-1B holders are not on that list. That is a reading of the text, not wording the statute uses, but it explains the case: an advertisement open only to H-1B, OPT or H-4 visa holders shuts out protected individuals.

The statute has limits. It does not apply to an employer of three or fewer employees, § 1324b(a)(2)(A). Citizenship-status restrictions are permitted where required to comply with law, regulation or executive order, or by a government contract, § 1324b(a)(2)(C)³.

What the settlement says

The Department's Immigrant and Employee Rights Section (IER) accepted a charge on 1 December 2025 under § 1324b(a)(1)(B). The agreement records IER's determination that there was "reasonable cause to believe that Respondent posted discriminatory advertisements in violation of 8 U.S.C. § 1324b"¹. The signature block shows 23 February 2026; the agreement takes effect on the date of the latest signature and runs for three years from then¹.

The AI element appears only in the press release of 25 February 2026 (number 26-190). It says the advertisements were "generated by an artificial intelligence (AI) tool" and restricted consideration to applicants holding H-1B, OPT or H-4 status. It does not name the tool. The Assistant Attorney General is quoted as saying the Department will not tolerate discriminating against U.S. workers "no matter who — or what — drafts a job advertisement"². The signed agreement contains no mention of artificial intelligence¹.

What compliance requires

The statute requires no citizenship-status restriction in recruitment unless an exception applies³. The agreement adds obligations for this respondent only¹:

• Review and revise employment policies on job advertising, recruiting and hiring within 60 days, covering citizenship status, immigration status and national origin.
• Submit any later policy change to IER for approval at least 30 days before it takes effect.
• Train within 60 days everyone who drafts or approves advertisements, screens applications or hires, and send attendance records to IER.
• Post IER's "If You Have The Right to Work" poster in English and Spanish within 14 days, including electronically where applicants complete hiring steps.
• Answer IER inquiries during the term, including written reports, premises inspection and interviews, and cure any alleged breach within 30 days of notice.

What enforcement has done with it

The civil penalty is $9,460, paid in two instalments of $4,730, the second due by 15 May 2026¹. The regulation sets the first-order penalty under § 1324b(g)(2)(B)(iv)(I) at $590 to $4,730 per individual discriminated against, for penalties assessed after 3 July 2025⁴. The total equals twice that maximum. That is an inference: the agreement does not say how the figure was reached, and I have not verified whether a later inflation adjustment applies.

The press release calls this the eighth settlement since the Department relaunched its Protecting U.S. Workers Initiative in 2025². The thirteenth, with OpenAI on 4 August 2026, was larger: $3.2 million, made up of $1.2 million in civil penalties and a $2 million back-pay fund. It concerned PERM recruitment practices, such as paper-only applications, and fewer than ten positions, not AI tools⁵. The two cases should not be read together as an AI enforcement line.

Where it does not reach

• The agreement imposes no AI-specific term: no tool review, vendor control or logging requirement¹.
• It binds the respondent alone. None of the sources opened reports an administrative law judge or court ruling on AI-drafted advertisements.
• The AI attribution is an allegation. The agreement contains no non-admission clause that I found, but it resolves differences only through the effective date and leaves IER free to investigate later conduct¹.
• It is a United States federal matter under § 1324b and says nothing about other jurisdictions.
• Whether the Department's emphasis on drafter-neutrality reflects the initiative's priorities or a particular concern about AI cannot be settled from this record. A reading that treats AI as incidental to a broader enforcement drive fits the same facts.

This brief is not legal advice.

What to do

Correct one assumption: that using a tool changes who answers for an advertisement. On this record it does not, because § 1324b applies to the person or entity that discriminates in recruitment, and nothing in it turns on who drafted the text³.

Then apply a bounded test. Before any advertisement goes out under your name, whether drafted by a tool, a recruiter or an agency, check that its wording does not limit consideration by citizenship or visa status, unless a law, regulation, executive order or government contract requires it. If you use AI drafting, put that check in the approval step. The settlement requires comparable training from this one respondent; for everyone else it is sensible practice, not a stated obligation.