Two job applicants say they were ranked by software on a scale they were never shown. The law firm Fisher Phillips, writing an analysis of the case for employers in January 2026, reports that Erin Kistler and Sruti Bhaumik filed suit against Eightfold AI in California state court on January 20, 2026, and that their complaint describes a platform that ranked candidates on a 0-5 scale based on their predicted likelihood of success in the role.
Whether that happened is for a court to decide. What the case exposes is easier to check, because it does not depend on the verdict. If a screening vendor does put a number on you, the ordinary hiring flow contains no step where you would be shown it.
What is a hiring score, and who assigns yours?
A hiring score is a rank that a vendor's software attaches to a candidate and sells to an employer as a way of sorting a pile that has grown faster than any recruiting team. The employer sees a sorted list. The candidate sees an application form and then silence.
What goes into the rank is the part candidates tend to get wrong, because they assume it reads the document they submitted. According to Fisher Phillips's account of the Eightfold complaint, the plaintiffs allege the platform "gathered information from third-party sources including LinkedIn, GitHub, Stack Overflow, and other public databases", analyzed "more than 1.5 billion global data points" including profiles of over 1 billion workers, and created inferences about applicants covering, in the complaint's words, their "preferences, characteristics, predispositions, behavior, attitudes, intelligence, abilities, and aptitudes". Eightfold responded in a media statement, quoted in the same Fisher Phillips analysis, saying that they "do not scrape social media and the like".
Those are allegations about one vendor. How many employers buy scoring of this kind, and how heavily they lean on it, is not something the complaint establishes and not something any figure in this post measures.
Why can you not see your own screening score?
Because a standard application flow contains no step where the score would be surfaced to the candidate. The scoring happens between a vendor and an employer. The candidate is the subject of the transaction without being a party to it, and a rejection email rarely explains what produced it.
That gap is what the Eightfold plaintiffs are pressing on. Fisher Phillips reports they claim they were never told a consumer report would be created, never authorized its creation, and "never had an opportunity to review or dispute the information before being rejected".
The practical effect is familiar to anyone reading rejections for a signal that is not in them. A score can sit between an application and a human reviewer and produce an outcome that looks exactly like ordinary silence. Four-Leaf's guide to why you are not getting interviews treats that silence as a funnel problem, and an invisible score is one more reason the funnel is hard to read from the candidate's side.
What does the Eightfold lawsuit actually allege?
The legal theory is narrower and more interesting than a discrimination claim. Per Fisher Phillips, the plaintiffs argue these assessments are "consumer reports" under the federal Fair Credit Reporting Act and California's Investigative Consumer Reporting Agencies Act, which would make the company assembling them a consumer reporting agency with disclosure, accuracy and dispute obligations attached.
The argument does not depend on proving the scores are biased. It depends on what the scores are, which is why Fisher Phillips flags the case as potentially reaching vendors and employers who would not consider themselves in the background-check business at all. The complaint also alleges the platform "provided these assessments to employers who used them to filter candidates before any human review", and that allegation is what makes the consumer-report framing plausible.
None of this has been decided. A complaint is one side's account, Eightfold disputes the characterization of its data sources, and Fisher Phillips itself notes the legal issues are novel enough that "the case could take years to resolve". That analysis is dated January 2026 and is the most recent view of the case reflected here, so the docket may have moved since. Anyone telling you that AI hiring scores are settled law is ahead of the record.
Would the FCRA give you a right to see your score?
Only if a court agrees the score is a consumer report, which is the open question rather than a settled one. The rights are worth knowing anyway, because they describe what disclosure would look like if the theory succeeds.
Where the FCRA does apply to employment screening, the Federal Trade Commission's guidance for employers sets out the sequence. The employer has to tell the applicant that information in a consumer report may be used for decisions about their employment, and get "written permission from the applicant or employee". Before rejecting someone on the strength of that report, the employer owes them "a notice that includes a copy of the consumer report you relied on to make your decision", plus a summary of the applicant's rights under the statute. The candidate gets to see the document and contest it.
Set that against a ranking that arrives with no notice and no copy, and the shape of the plaintiffs' argument is clear enough. Whether it holds is for the courts.
Why is a private lawsuit carrying this question?
Federal guidance once answered a version of it directly. The Consumer Financial Protection Bureau issued Circular 2024-06, titled Background Dossiers and Algorithmic Scores for Hiring, Promotion, and Other Employment Decisions and published at 89 FR 88875. The circular asked whether an employer could make employment decisions using "background dossiers, algorithmic scores, and other third-party consumer reports about workers" without adhering to the Fair Credit Reporting Act. Its response was no. The reasoning was that such dossiers, when obtained from third parties and "used by employers to make hiring, promotion, reassignment, or retention decisions are often governed by the FCRA".
The third-party condition in that sentence carries weight, because a score an employer builds in house sits outside the reasoning. A circular is also a policy statement aimed at enforcers rather than a rule binding employers.
The Bureau then withdrew it. A Federal Register notice retiring a large batch of Bureau guidance lists that circular among the withdrawn documents and states that the "withdrawals are applicable as of May 12, 2025". The statute did not change and the agency's published reading of it was pulled back, which leaves private litigation as the mechanism actually testing the question.
What is overrated
Optimizing your resume to beat the bot. The instinct is understandable and it aims at the wrong target, because a rank that a vendor allegedly assembles from third-party profile data is not something a keyword-stuffed resume moves.
The auto-reject story is also weaker than its popularity suggests. Enhancv conducted 25 in-depth interviews with U.S.-based recruiters and talent acquisition professionals between September and October 2025, covering more than 10 ATS platforms, and reported that "Of those recruiters, 23 (92%) said their systems do not auto-reject resumes for formatting, content, or design." Enhancv sells resume optimization, so the finding cuts against its own commercial interest, and 25 recruiters is a small, non-representative panel. Four-Leaf's explainer on what an ATS actually does covers the parsing and ranking mechanics in detail. Clean parsing is table stakes, and the automated filters that do fire are mostly knockout eligibility questions.
What to do this week
- Score the document you control. Four-Leaf's free resume checker returns a 0-100 ATS readability and writing score with the specific fixes behind it, without an account or a job description. It reads the resume you paste or upload, plus an optional target job title, and it never sees the job posting, so treat the number as a measure of your document and not as a preview of any employer's internal ranking.
- Ask at the recruiter screen whether an automated assessment is part of the process. Phrase it as curiosity about the timeline. The answer tells you where the decision is being made, and a recruiter who cannot say is itself informative.
- Make your public profiles consistent with your resume. If inference layers read public professional data, as the Eightfold complaint alleges, contradictions between your LinkedIn history and your resume are cheap to fix and bad to leave.
- Read rejections by stage instead of by wording. A rejection at hour six and a rejection after an onsite carry different information. How long it takes to hear back after applying sets the baseline timings worth comparing yours against.
- Do not build your plan around getting your file. Asking is reasonable. Counting on an answer, while the FCRA question sits in front of a court, is not.
Where this is heading
The Eightfold case matters beyond its own verdict, because the question it raises outlives it. Scoring layers can sit between a candidate and a human reviewer while staying invisible to the person being scored, and opacity of that kind usually ends through litigation, legislation or a vendor deciding that transparency sells better.
Until one of those lands, the asymmetry is the working condition. Someone may hold a number about you that you cannot read, built partly from data you did not submit, and the only scores you can inspect are the ones you run yourself. That is a thin form of control. It is also the kind on offer, which is a good reason to use it.