1. State v. Loomis (Wisconsin Supreme Court, 2016)

This is the most cited case on algorithmic sentencing tools.

The defendant was sentenced partly using the COMPAS algorithm, a proprietary risk assessment tool used to predict recidivism.

Key issue:

Whether using a black-box algorithm in sentencing violates due process.

Holding:

The court allowed use of COMPAS but imposed caution.

Legal significance for traceability:

• The court acknowledged lack of transparency in the algorithm
• Defendants could not fully challenge how the score was computed
• Court required warnings about algorithmic limitations

Why it matters:

It is a foundational case showing that U.S. courts may allow AI decisions even when the reasoning is not fully traceable or explainable, but with procedural safeguards.

2. Mobley v. Workday, Inc. (N.D. California, 2024)

A major modern AI hiring discrimination case involving ML-based screening systems.

Key issue:

Whether AI hiring tools can be liable under Title VII for discriminatory outcomes.

Allegation:

Workday’s AI-driven applicant screening system allegedly rejected applicants based on protected characteristics (age, race, disability).

Holding (procedural stage):

The court allowed the case to proceed past motion to dismiss.

Traceability relevance:

• Plaintiff sought discovery of how the ML system made decisions
• Raised issue of algorithmic opacity in hiring
• Court recognized AI vendor may be treated as an “agent” of employer

Why it matters:

This case pushes toward forced algorithmic traceability through discovery in employment discrimination cases.

3. EEOC v. iTutorGroup, Inc. (E.D.N.Y., 2023)

One of the first EEOC cases explicitly involving AI screening.

Key issue:

Age discrimination in automated hiring systems.

Facts:

AI system allegedly rejected applicants based on age-related data inputs.

Outcome:

Consent decree settlement required compensation and injunctive reforms.

Traceability relevance:

• EEOC obtained access to algorithmic hiring logic through enforcement
• Company required to change or monitor automated decision systems
• Demonstrates regulatory push for auditability of ML systems

Why it matters:

Shows government forcing internal visibility into ML decision-making pipelines.

4. Syed v. M-I, LLC (9th Cir. 2020)

A Fair Credit Reporting Act (FCRA) case involving automated background screening.

Key issue:

Whether employers must disclose third-party automated background check systems.

Holding:

The court held that failure to clearly disclose the nature of the background report process can violate FCRA.

Traceability relevance:

• Employees must be informed when automated systems influence hiring decisions
• Limits “black box” screening without disclosure
• Strengthens right to understand algorithmic inputs affecting employment

Why it matters:

Establishes procedural transparency requirements for automated decision systems.

5. HiQ Labs v. LinkedIn Corp. (9th Cir. 2019; later proceedings 2022)

A major case about data access for algorithmic analysis.

Key issue:

Whether scraping publicly available LinkedIn data is unlawful under the Computer Fraud and Abuse Act (CFAA).

Holding:

Public data scraping was largely allowed under certain conditions.

Traceability relevance:

• HiQ used scraped data to build predictive analytics models
• Court limited LinkedIn’s ability to block access to public data
• Supports indirect ML traceability by enabling external auditing of platforms

Why it matters:

This case supports the idea that algorithmic systems can be externally analyzed using public data, limiting absolute opacity.

6. FTC v. Everalbum, Inc. (2021)

A Federal Trade Commission enforcement action involving facial recognition ML systems.

Key issue:

Misleading users about use of facial recognition technology and data retention.

Outcome:

Company required to delete certain ML models trained on improperly retained biometric data.

Traceability relevance:

• Required deletion of trained AI models derived from unlawful data
• FTC enforced transparency and deletion obligations
• Recognized ML models as legally relevant “derivative data products”

Why it matters:

Shows regulatory recognition that ML models themselves must be traceable to lawful data sources.

7. (Bonus contextual case) Algorithmic accountability trend in U.S. courts

Across these cases, courts increasingly recognize:

A. Limited “black box tolerance”

(Allowed in Loomis, but with warnings)

B. Expanding discovery rights

(Mobley v Workday, EEOC cases)

C. Mandatory disclosure duties

(Syed v M-I, iTutorGroup)

D. External auditing via data access

(HiQ v LinkedIn)

E. Regulatory enforcement of model provenance

(FTC v Everalbum)

Overall Legal Meaning of “ML Traceability Rights” in the U.S.

While not a single statutory right, U.S. law is evolving toward a bundle of enforceable rights:

• Right to contest automated decisions
• Right to obtain meaningful disclosure (limited explainability)
• Right to discovery of algorithmic inputs in litigation
• Right to non-discriminatory algorithmic outcomes
• Regulatory oversight of training data and model behavior

However, a major tension remains:

Courts often balance transparency against trade secret protections, meaning full algorithmic disclosure is still rare.