Domain Atlas / Hiring & employment screening AI
SiriusXM's iCIMS applicant screening
Explore this deployment in the PAN Lab ↗
In the PAN Lab, the readouts of this case's model organization carry a shaded evidence band whose width follows the least-established class among the modeling inputs the readings rest on.
The least-established input behind this case's model organization's readings is an assumption, not a measurement. Evidence base: 4 assumed · 16 published baseline.
Arshon Harper, an African-American information-technology professional in Detroit, applied for approximately 150 positions with Sirius XM Radio, LLC through the company's iCIMS-powered application platform between November 2023 and 21 November 2024, and was rejected for all but one. The single exception was a thirty-minute interview for an IT Desktop Support role in late 2023, which also ended in rejection, so the pleaded record is zero offers from about 150 applications and 149 without an interview. The complaint states that arithmetic as a 99.3 percent rejection rate and offers it as its pattern evidence. His pleaded qualification profile is a 2019 B.S. in Business Administration from Wayne State University and more than ten years of information-technology experience, including Tier 2 support to about 4,500 employees at Wayne State's computing and information technology unit between 2018 and 2021 and data-collection work at the Detroit Department of Transportation since 2008; his named targets were IT Desktop Support, Software Engineer and Technical Support Specialist. The employer is pleaded to receive thousands of applications annually. Every figure here is a pleading. It is one applicant's personal rejection rate rather than a measured disparity across applicants, no class-wide or agency-collected data exists in this record, and whether an individual probe record can support a systemic claim is precisely what the pending motion for judgment on the pleadings puts in issue.[2]
What happened
Arshon Harper applied for a job at Sirius XM Radio in November 2023. Then he applied for another, and another. By 21 November 2024 he had applied for approximately 150 positions with the company — IT Desktop Support, Software Engineer, Technical Support Specialist — through the careers platform Sirius XM runs on iCIMS, a commercial applicant tracking system. He was rejected for all but one of them. The exception was a thirty-minute interview for a desktop-support role in late 2023, and it also ended in rejection. The complaint he later filed states that record as a 99.3 percent rejection rate and offers it as its pattern evidence.
His pleaded qualifications are on the record too: a 2019 B.S. in Business Administration from Wayne State University, more than ten years in information technology including Tier 2 support to about 4,500 employees at Wayne State's own computing unit between 2018 and 2021, and data-collection work at the Detroit Department of Transportation since 2008. The degree is in business administration rather than computer science, which the pleading itself sets out as part of the profile.
What the complaint says happened in between is an allegation, and most of it is pleaded expressly on information and belief. It alleges that Sirius XM, by and through the iCIMS platform, uses AI and machine-learning candidate matching, shortlisting and sourcing features that evaluate applicants "based on data points (e.g., educational institutions, employment history, zip codes) that proxy for race" — a triple stated in the introduction, repeated at paragraph 20 and carried into two of the three counts. It alleges the tools were built on historical hiring data, carrying decades of earlier human selection into an automated screen and applying it at scale. It alleges the tools intentionally and disproportionately reject African-American applicants, and that they are not job-related and lack business necessity. Sirius XM answered on 6 January 2026 denying liability with affirmative defenses.
Part of the pipeline is not in dispute in the same way, because the complaint pleads it from the vendor's own published product documentation rather than from any allegation about this employer. A submitted resume is parsed on arrival; name, contact information, skills, work history and education are extracted; and the platform generates its own list of the applicant's skills from the full resume text rather than using the list the applicant wrote. That generated list is what recruiters search when they source candidates. So before any judgement is exercised about a person, a description of that person has been written by the system and substituted for their own.
There is one more mechanism pleaded, and it is unusual enough to be worth reading slowly. When a Sirius XM official questioned Harper's use of multiple email addresses, he explained that he had used them to circumvent what he suspected were algorithmic penalties for repeat applications — and he pleads that this circumvention is what produced his only interview. That is a screened-out person probing an opaque system by perturbing his own inputs, and it is the only observation of the alleged applicant-history mechanism anywhere in the record. Whether such a mechanism exists has not been discovered, let alone adjudicated.
The governance channels arrive next, and neither of them has evaluated the tool. Harper filed an EEOC charge on 22 November 2024. On 6 May 2025 the Commission's Newark Area Office issued a Determination and Notice of Rights stating that it "will not proceed further with its investigation and makes no determination about whether further investigation would establish violations." About five and a half months from charge to closure, and the administrative gate opened the courthouse door without anyone ever examining the screening tool. The class action followed on 4 August 2025 in the Eastern District of Michigan, pleading Title VII disparate treatment, Title VII disparate impact and intentional race discrimination under 42 U.S.C. 1981, and seeking a declaration, a permanent injunction "prohibiting continued discrimination and requiring reforms," backpay, front pay, and compensatory and punitive damages.
The procedural texture is part of this record rather than an aside. Harper signed the complaint himself, with Winston Cooks, LLC — an Alabama employment-discrimination firm — listed of counsel and signing the civil cover sheet. The $405 filing fee went unpaid until 1 October 2025. Plaintiff-side attorney appearances came on 26 September and 8 October 2025; defense appearances followed on 4 and 5 November. On 6 January 2026 Sirius XM answered with affirmative defenses and moved for judgment on the pleadings under Rule 12(c) on the same day. The motion was fully briefed on 12 March 2026 and noticed for determination without oral argument. On the docket record verified 28 August 2026, it remains undecided, there is no class-certification motion and there is no settlement.
iCIMS is not a party. Its public posture is a formal responsible-AI program built on six pillars — human-led, transparent, private and secure, inclusive and fair, technically robust and safe, and accountable — describing processes for bias audits, transparency reporting and human oversight that reference New York City Local Law 144's automated-employment-decision-tool obligations, alignment claims to the National Institute of Standards and Technology's AI Risk Management Framework, the OECD AI Principles and ISO 42001, and a TrustArc TRUSTe Responsible AI certification it says it obtained in March 2025 and claims as a first among enterprise recruiting software providers. All of that is the vendor's own account of its own product line. It discloses no audit results, and nothing in the record connects it to what Sirius XM actually configured or ran. No employer-side bias audit, validation study or monitoring of this configuration appears anywhere in the public record reviewed. No parallel litigation over this platform and no enforcement action against the vendor was identified, which is stated as none identified rather than none existing.
The sociotechnical reading
Start with what is unusual about who is answering. In the vendor-side litigation that this case is repeatedly read beside, a screening platform is the defendant on the theory that it acts as the employers' agent — so the governance pressure lands on the platform seam, one tool serving thousands of employers at once. Here the platform is not sued at all. The employer stands alone for the outcomes a licensed tool is alleged to produce, and the proposition independent analysts say the case tests is the ordinary one: responsibility under federal anti-discrimination law does not travel to the vendor along with the technology. Same domain question — who owns the screen — answered from the opposite side.
That answer determines the shape of everything else. There is one employer's pipeline rather than a market's; a single-firm class rather than a cross-employer collective; race under two statutes rather than age under one; and a vendor whose only appearance in the record is its own published description of its own controls. It also determines where the humans sit. As pleaded, recruiters see what the matcher surfaces, and 149 of about 150 applications reached no human decision point at all. That is not a story about people deferring to a machine's recommendation. It is a story about a stage that most applications never reach, and it means the screen's rejection was the decision rather than an input to one.
Two absences follow, and they are the load-bearing ones. First, nothing reconciles an automated rejection against the file it was computed from before it goes out. No source describes a step of that kind, and no channel exists for an applicant to contest a screening rejection short of filing a federal charge — which is what happened here. Second, nothing has reconciled the case record against the employer's own tracking record either. That is what discovery is for, and discovery has not opened; the motion under submission tests the sufficiency of the pleadings rather than the facts. Everything anyone outside the company knows about this screening configuration is a one-sided pleading, an answer denying it, a fully briefed motion, and an agency notice that expressly declines to say anything about the merits.
The evidence base is itself the case's structural question. The statistical showing is one applicant's record across about 150 applications, not class-wide or EEO-1 data. That is pleaded pattern evidence, and whether an individual probe record can carry a systemic pleading past Rule 12(c) is precisely what the pending motion asks. Read alongside it, the multiple-email-address episode is the rarest thing in this file: an applicant-side detection channel, operated by the person the screen was acting on, against a system he could not see. It worked once, and its result is the only observation of the alleged repeat-application penalty in the entire record.
The administrative channel deserves its own line, because it is easy to mistake for oversight. A charge was filed, an agency received it, and about five and a half months later the agency said in terms that it makes no determination. Functionally that is a pass-through to court, not a merits check — the one public body with subject-matter jurisdiction over the deployment looked at it and recorded nothing about it. Set beside a court that has ruled on nothing and a vendor programme that discloses no results, the honest reading is that this deployment has never been examined by anybody.
Two boundaries hold across the whole reading. Applicants are the screened population and are not modelled as part of the deployment's dynamics: no screening outcome, ranking, rejection or employment consequence for any person is computed from the network drawn here, and the applicant enters it only as the thin transfer by which one person's own file became the case record's pattern evidence. And the register is allegation throughout for everything about how the screening behaves. What can be asserted as fact is the procedural history, the counts pleaded, the class sought, the relief sought, the charge date and the agency's express no-determination — and the vendor's claims, labelled as the vendor's. Everything else is what one complaint says, in a case where nothing at all has been decided.
The concepts used in this reading are defined in the Field Guide; the governance responses live in the Practice Library. The model organization for this case can be stress-tested in the PAN Lab.