Two federal lawsuits that accused employers of using screening technology to reject applicants unfairly were dismissed within a day of each other this fall. The first was against Sirius XM in Michigan on September 30, 2026. The second was against the insurer Assurant in Alabama the next day.
Headlines about cases like these tend to read as "employer wins, AI cleared." That is not what happened. Neither court decided that the screening tools were fair or lawful. Both courts decided that the applicant had not yet put enough facts in the complaint. The difference matters to both sides of the hiring table, because it shows exactly what evidence a court expects to see, and that is the same evidence a careful employer should be able to produce about its own process.
Harper v. Sirius XM: 150 Applications, One Interview
Arshon Harper, a Black information technology professional from Detroit, said he applied for about 150 jobs at Sirius XM between November 2023 and November 2024, according to HRMorning's October 27, 2026 report. All but one application was rejected before the interview stage. He had one 30-minute interview and was rejected after it.
He claimed that Sirius XM used artificial intelligence features in its iCIMS applicant tracking system (ATS) to score candidates, and that those features relied on stand-ins for race such as schools attended, work history, and ZIP codes. He sued under Title VII and Section 1981 and asked to proceed as a class action. Sirius XM said it does not use artificial intelligence in hiring. HRMorning reports that the court set that question aside and assumed, for purposes of the ruling, that it did.
Judge Terrence Berg still dismissed the complaint. According to HRMorning, Harper alleged no facts suggesting race motivated the rejections, such as losing the job he interviewed for to a non-Black candidate. He also supplied no group-level statistics, no comparison group, and no examples of other applicants with similar experiences. His argument that repeated rejections plus the company's ability to collect demographic data were enough did not persuade the court. In the court's words, as quoted by both Fisher Phillips and HRMorning, "data collection is not the same thing as data discrimination."
The dismissal was without prejudice, which means the case was dismissed for now and not decided on the merits.
Knight-Bell v. Assurant: The Keywords Were Chosen by People
The second case, as Fisher Phillips describes it in its October 27, 2026 analysis, involved Vanessa Knight-Bell, an African American woman over the age of 40. She alleged that Assurant's artificial intelligence hiring tools screened out all 21 of her applications because of her race and age.
Assurant denied using artificial intelligence to filter resumes. It described a process in which managers ran keyword searches to cut hundreds of applications for a posting down to a manageable number. Judge Harold Mooty found that Knight-Bell had identified a specific practice that could support a claim: employer-selected keywords deciding which applications got further review. But Fisher Phillips reports that her complaint did not supply any actual search term or other facts making it plausible that the practice disproportionately excluded people in her protected group.
She tried to fill the gap with academic research about bias in machine learning models. The court pointed out that her complaint actually targeted a human manager's choice of search terms, not artificial intelligence. Her age claim also failed. The court followed Eleventh Circuit precedent holding that the disparate impact provision of the Age Discrimination in Employment Act does not cover job applicants.
That last detail is worth a second look. A keyword list typed by a recruiter is a screening practice, just as an algorithm is. Fisher Phillips makes the point directly for employers: "Human discretion does not categorically remove a practice from disparate impact scrutiny."
What a Dismissal Does and Does Not Mean
Fisher Phillips draws three common threads from the two decisions. The first is stated plainly: "Repeated rejection alone does not establish group bias." The second is that a claim that technology could use a biased proxy is not the same as facts showing that it did. The third is that intentional discrimination and disparate impact are separate theories with separate proof.
Read carefully, these rulings set a pleading bar. They do not say:
- That the screening tools in either case were unbiased.
- That employers can ignore how their ATS settings and keyword lists affect different groups.
- That a future applicant with group-level data could not bring a stronger case.
Both courts told future plaintiffs, in effect, what to bring next time: the specific practice, the specific criteria, and facts showing a disproportionate effect on a group. Employers should expect better-prepared complaints, not fewer of them.
For Job Seekers: Keep Your Own Records
If you suspect a screening process is filtering you out unfairly, these cases show that a personal rejection count, even a long one, is not enough on its own. What helps is a record that someone else can check later.
- Log every application. Record the date, the job title, the posting link, and the outcome. Save a copy of the posting itself, because postings are often taken down.
- Note the stated requirements. If you met every listed requirement and were rejected within minutes, write that down with the timestamps.
- Watch for knockout questions. Many automatic rejections come from screening questions, not resume scoring. Our post on knockout questions that reject qualified candidates explains how they work.
- Check how your resume reads to a keyword filter. In the Assurant case the screening was a keyword search. A resume that uses different words for the same skill can miss a keyword list entirely. Our post on why keyword matching is not hiring intelligence covers the gap.
This is general information, not legal advice. If you believe you were treated unfairly, an employment attorney or the agency that handles discrimination charges where you live can tell you what applies.
For Recruiters and HR Teams: Know What Your Process Does
The first of Fisher Phillips' five recommended steps for employers is "Identify what your hiring system actually does." That includes features switched on by default, screening criteria, and the points where a person reviews the results. The steps below follow from both decisions.
- Inventory every filter. List each knockout question, auto-reject rule, score threshold, and saved keyword search. Our ATS filter audit checklist walks through the inventory.
- Review the keyword lists people build. Check whether each search term is tied to the actual job. A term that is a habit rather than a requirement is the kind of practice the Assurant court said could support a claim if a disparity were shown.
- Look at your own outcome data. Fisher Phillips recommends a bias audit even where no law requires one, conducted at the direction of counsel. Our post on what an AI hiring bias audit should contain describes the parts of a useful one.
- Keep screening records. Be ready to explain how each applicant was screened. If you hire in California, our post on the four-year record rule covers how long to keep them.
An employer that can show which criteria were used, why each one relates to the job, and what the outcomes looked like across groups is in a far better position than one that learns how its own filters work from a complaint.
Where TrueScan HR Fits
TrueScan HR scores a resume against the job description and shows the matched and missing skills behind every score. A recruiter can see why a candidate scored the way they did, and a job seeker can see which requirements a resume does not show clearly. That visibility is the opposite of a keyword list nobody remembers writing.
The Bottom Line
The Sirius XM and Assurant dismissals are wins for those two employers at an early stage, not a verdict on screening technology. The courts asked for specific practices, specific criteria, and group-level facts. Job seekers who keep records and employers who can explain their own filters will both be better prepared for the next case, which is likely to arrive with more evidence than these two did.
Sources: