Recent Employment Discrimination Case Offers Lessons on AI and Keyword Bias Claims

PDF

The appeal is undeniable. Algorithms, automated decision systems, and artificial intelligence (AI) can increase efficiency for many employers. The technology can streamline the processes for hiring, performance evaluations, compensation decisions, and even terminations. However, with the increased use and reliance on such decision-making software in the employment context come legal risks. Discrimination claims can arise under Title VII, the Age Discrimination in Employment Act, the Americans with Disabilities Act, and other federal and state laws. A recent case in federal court serves as an important reminder.

In the case, brought in 2025, an African-American male applicant filed an employment discrimination lawsuit in a Michigan federal court against Sirius XM, a provider of satellite radio and other streaming services. In his complaint, Arshon Harper alleged that he had applied to about 150 positions but that Sirius had rejected all but one of his applications before the interview stage. He was not hired for the one position for which he interviewed. Mr. Harper alleged that Sirius used a third-party vendor’s applicant tracking system, including AI, to screen candidates and that the AI technology screened him out because of his race. More specifically, according to Mr. Harper, those AI tools evaluated applicants using data points that acted as unlawful proxies for race, including educational institutions, employment history, and zip codes, which resulted in the disproportionate rejection of qualified African-American candidates in comparison to similarly qualified non-African-American candidates. Earlier this year, Sirius moved to dismiss the complaint for failure to state a claim. In response, Mr. Harper argued that the number of rejections he had received—coupled with his allegation that Sirius was able to collect and use demographic data through the applicant tracking system—was sufficient proof that Sirius could plausibly have used racial characteristics to discriminate against him.

On September 30, 2026, the court rejected Mr. Harper’s arguments and dismissed his claims, ruling that the complaint failed to point to direct evidence sufficient to demonstrate intentional discrimination. In its reasoning, the court noted that “data collection is not the same thing as data discrimination” and held that Mr. Harper’s repeated rejections and the collection of data were insufficient to establish claims of disparate impact or disparate treatment. “In fact,” the court noted, “data collection can be an essential tool for a company to ensure its otherwise neutral policies and practices do not disproportionately disadvantage applicants based on a protected class status” (emphasis in original). The court also found that Mr. Harper had failed to plead factual allegations regarding comparators.

While the court ultimately rejected the claims in the lawsuit, the case is a reminder that employers must exercise caution when using AI-powered employment screening tools. Among the key takeaways, employers using AI for hiring should make certain that they know and understand what the AI-powered system does in terms of tracking, searching, and screening. Employers using a third-party vendor should work with that vendor to understand what features are enabled, what search terms are used, and when to have a human involved in the process. Also, employers should make sure that the tracking, search terms, and screenings are tailored to the specific job positions, and they should document how the decisions were made to use the various features. Finally, if an applicant brings a claim, employers should consult legal counsel to investigate whether the challenged process can be connected to discrimination.