State v. Miles: Case-Specific Facial Recognition Discovery and a Deferred, Burden-Shifting Path to Proprietary Source Code

1. Introduction

State v. Tybear Miles addresses how much discovery a criminal defendant may obtain when the State used Facial Recognition Technology (FRT) during an investigation. A grand jury charged Tybear Miles with first-degree murder and weapons offenses arising from a fatal shooting in Jersey City. A confidential informant (CI) viewed surveillance footage from near (but not depicting) the shooting, identified two individuals by nicknames, and provided Instagram usernames. Police ran an FRT search on an Instagram profile photo associated with one nickname; the search returned defendant among multiple potential matches. Police later obtained identifications from defendant’s sister, ex-girlfriend, and others based on additional surveillance images; no one identified defendant as the shooter, and no video captured the shooting.

The key dispute was whether the trial court could compel production of the same “thirteen items” of FRT discovery previously ordered in State v. Arteaga, 476 N.J. Super. 36 (App. Div. 2023)—including proprietary information such as source code—simply because FRT was used. The Supreme Court rejected that mechanical approach and announced a case-specific framework: baseline, non-proprietary disclosures are generally required, while proprietary disclosures (like source code) must await a developed record and a showing of particularized need.

2. Summary of the Opinion

3. Analysis

3.1 Precedents Cited

A. Foundational defense and due process principles

• State v. Garron, 177 N.J. 147 (2003) (quoting Crane v. Kentucky, 476 U.S. 683 (1986)): The Court framed discovery as instrumental to the constitutional right to “a meaningful opportunity to present a complete defense.” This principle supported requiring disclosure of how FRT was used so the defense can test identifications, investigative choices, and alternative-suspect theories.
• Brady v. Maryland, 373 U.S. 83 (1963): Cited for the State’s obligation to disclose favorable, material evidence. The Court used Brady to emphasize that discovery duties reach beyond what the State “plans” to use at trial; information about investigative steps can still be material to guilt, impeachment, or defense theory.
• United States v. Bagley, 473 U.S. 667 (1985) and State v. Brown, 236 N.J. 497 (2019): Invoked to underscore that impeachment evidence is within Brady. That mattered here because FRT use could bear on impeachment of investigative decisions and of identification witnesses whose exposure to police-selected images may have been influenced—directly or indirectly—by the FRT lead.

B. New Jersey’s broad criminal discovery regime and relevance as the “touchstone”

• State v. Hernandez, 225 N.J. 451 (2016): Provided the abuse-of-discretion review standard and the proposition that trial courts may order discovery beyond the rules to ensure fairness and truth-seeking. The Court relied on Hernandez to justify requiring baseline FRT-related disclosures even while rejecting an overbroad, checklist-driven order.
• State v. Ramirez, 252 N.J. 277 (2022) (quoting State v. S.N., 231 N.J. 497 (2018)): Used to define when discovery rulings are an abuse of discretion—e.g., failing to consider relevant factors or misapplying law. This framed the partial reversal: the trial judge erred by compelling proprietary disclosures without conducting the required, fact-specific inquiry.
• State v. Scoles, 214 N.J. 236 (2013): Reinforced that New Jersey favors “automatic and broad discovery” supporting an “open-file” approach. That policy backdrop helped the Court define “minimum necessary” disclosures when FRT has been used.
• State v. Cook, 43 N.J. 560 (1965) and State in Int. of W.C., 85 N.J. 218 (1981): Cited for discovery’s truth-seeking function and the court’s inherent power to order discovery when justice requires. These supported the Court’s insistence that defendants receive enough information to meaningfully contest FRT-driven investigative pathways.
• State v. Desir, 245 N.J. 179 (2021) (quoting State v. Richardson, 452 N.J. Super. 124 (App. Div. 2017)): Provided the definition of “relevant” material (“tendency in reason to prove or disprove a fact of consequence”). The Court used relevance to anchor its two required discovery categories: identifying the tools (to assess reliability) and explaining their use (to test investigation/identification/third-party guilt).
• State v. D.R.H., 127 N.J. 249 (1992): Served as a limiting principle against “unfocused, haphazard” fishing expeditions. The Court’s rejection of a rigid “thirteen-item checklist” reflects this: discovery must be tethered to case needs, not categorical demands.
• State in Int. of A.B., 219 N.J. 542 (2014): Reinforced the court’s ability to order discovery beyond rule minimums when needed to ensure fairness. This supported baseline FRT disclosures even where the State argued it would not introduce FRT at trial.

C. Proprietary technology discovery: the governing framework

• State v. Pickett, 466 N.J. Super. 270 (App. Div. 2021): Central to the Court’s treatment of proprietary FRT information. Pickett recognized that, “to the extent necessary to ensure a fair trial,” defendants may obtain proprietary software information, potentially including source code, but only through a burden-shifting, fact-intensive analysis: first, the party resisting discovery shows good cause (confidential/proprietary); then the defendant must show “a particularized need,” guided by considerations like expert support, specificity, protective orders, and case-unique factors. The Supreme Court applied Pickett as the controlling template for proprietary FRT discovery and held the trial court bypassed that required inquiry.
• State v. Arteaga, 476 N.J. Super. 36 (App. Div. 2023): The key comparator case. In Arteaga, the Appellate Division ordered production of thirteen specific categories of FRT discovery because, “on the facts of this case,” the defendant demonstrated a “particularized need” under Pickett—supported by expert input and tethered to defense needs. The Supreme Court clarified that Arteaga does not create a universal “thirteen-item checklist” applicable whenever FRT appears in an investigation; rather, it exemplifies case-specific application of Pickett and relevance principles.
• Frye v. United States, 293 F. 1013 (D.C. Cir. 1923): Appears through Pickett as the context in which reliability of novel scientific evidence may be litigated. While the Court did not decide a Frye-type reliability hearing here, the opinion treats tool-identification discovery as a necessary predicate to develop a record that could later support reliability challenges.

3.2 Legal Reasoning

A. Relevance and fairness drive baseline FRT discovery

The Court reaffirmed that New Jersey criminal discovery is broad, but bounded by relevance. It then tailored a practical baseline: when FRT is used, defendants are ordinarily entitled (at minimum) to information (i) identifying the FRT tools/materials and (ii) explaining how they were used. This is not framed as a mere administrative preference; it is linked to the constitutional right to present a meaningful defense and to New Jersey’s open-file discovery culture.

B. Why “tool identification” is discoverable

The first category—identifying the FRT tools and materials—is justified as necessary to develop a record about reliability in the specific case. The Court described “basic information” (e.g., name/manufacturer, version numbers, years developed, key performance metrics including publicly available error-rate documentation, and database/operator information). Such disclosure allows the defense to evaluate what system was used and whether known limitations, performance constraints, or operational settings may affect the probative value of FRT outputs in that investigation.

C. Why “how it was used” is discoverable

The second category—how the State used FRT—includes the original “probe” image, edited versions, matched images, and related operational outputs. The Court found potential relevance to at least three defense functions:

• Impeaching identification evidence: Even if no witness identified defendant as the shooter, the defense may need to explore whether FRT-derived suspicion influenced what images were shown to interviewees and how identifications were obtained or reinforced.
• Challenging the investigation: The defense is entitled to test the thoroughness and neutrality of investigative steps, including whether FRT contributed to “tunnel vision,” what officers understood about the tool’s limitations, and whether alternative leads were pursued.
• Supporting third-party guilt: The Court highlighted that the FRT searches returned multiple “matches,” including instances where defendant was not ranked as the closest match. Those results can be relevant to alternative-suspect theory and to contest the strength of the State’s path to defendant.

D. Why proprietary source code was not compelled (yet)

The Court’s partial reversal is anchored in Pickett’s burden-shifting framework. The trial judge compelled the Arteaga list wholesale, without first: (1) requiring a good-cause showing that requested items were proprietary/confidential; and (2) requiring defendant to demonstrate “particularized need.”

The Supreme Court also identified two practical, record-based reasons to defer the proprietary question: (i) the record was not developed enough to assess reliability questions about the “particular FRT used” (as contrasted with generalized concerns); and (ii) defense counsel acknowledged uncertainty about whether source code would be relevant or helpful. Accordingly, the Court reversed the proprietary portion “without prejudice” and invited a renewed request on remand if the defense can make the required showing.

3.3 Impact

A. The new organizing principle: no universal “Arteaga checklist”

The opinion’s most significant doctrinal contribution is its explicit rejection of a mechanical, one-size-fits-all adoption of State v. Arteaga’s thirteen-item list. Courts must instead tailor discovery to the case, using relevance and fairness as the guiding standards. This reduces the risk that trial courts treat Arteaga as an automatic entitlement to broad proprietary disclosures whenever FRT appears.

B. Establishing a “minimum necessary” baseline for most FRT cases

Even as it rejected a rigid checklist, the Court signaled that the two categories it mandated—(1) identifying the tools and (2) explaining their use—will “in most cases” constitute the minimum necessary to safeguard a fair trial. Practically, this creates a stable floor of FRT transparency in New Jersey criminal cases, making it harder for the State to characterize FRT as a purely internal lead immune from disclosure.

C. Proprietary technology discovery becomes more structured and staged

The Court’s insistence on a developed record before compelling source code channels future litigation into a staged approach: defendants first obtain basic tool-and-use information; only then, if warranted by what that discovery reveals (and supported by expert proffers), defendants may seek proprietary materials under Pickett. This approach is likely to:

• reduce premature source-code litigation;
• encourage targeted expert declarations tied to the actual system used and its operational settings;
• increase use of protective orders and in camera procedures to balance fair-trial needs with trade secret concerns.

D. Effects on policing and disclosure practices

By treating “how FRT was used” as discoverable, the opinion pressures agencies to document: probe-image provenance and edits, candidate lists, ranking outputs, analyst notes, and operator training/qualifications. Poor recordkeeping (as suggested by uncertainty in the record about where search results came from) becomes a litigation risk, with potential remedies left to trial-court discretion on remand if the State fails to produce required discovery.

4. Complex Concepts Simplified

• Facial Recognition Technology (FRT): Software that compares a “probe” photo to a database of photos and returns a ranked list of similar faces (“candidate list”). It does not necessarily declare identity; it typically returns similarity scores/rankings that require human interpretation and follow-up investigation.
• “Probe” photo: The image submitted into the system to search for similar faces in the database (here, an Instagram profile photo saved as a JPEG).
• Candidate list / match ranking: The ordered results returned by FRT, often showing multiple potential “matches,” with the system ranking or scoring them by similarity. A lower rank (e.g., 8th out of 10) can matter to the defense in arguing uncertainty or alternative suspects.
• Proprietary information / source code: Protected intellectual property describing how the software works internally (the “recipe” for the algorithm). Courts treat this differently from ordinary discovery because of trade-secret and security concerns.
• Burden-shifting and “particularized need” (from State v. Pickett): If the State shows good cause that requested material is proprietary/confidential, the defendant must then show a specific, case-tied need for that material—not a general interest in auditing the tool.
• “Without prejudice” reversal: The Court is not permanently denying proprietary discovery; it is saying “not yet,” leaving the door open to renew the request later on a stronger factual record.

5. Conclusion

State v. Tybear Miles establishes that FRT discovery in New Jersey is case-specific and cannot be reduced to a universal “Arteaga checklist.” At the same time, it cements a practical baseline: when the State uses FRT, it must generally disclose (1) what FRT tools/materials were used and (2) how they were used in the investigation—because those facts can bear on reliability, impeachment, investigative integrity, and third-party guilt.

The decision’s balancing move is equally important: proprietary FRT materials like source code are not categorically discoverable at the outset. Instead, courts must apply State v. Pickett’s burden-shifting framework on a more developed record, requiring a defendant to demonstrate particularized need. The resulting rule advances transparency and fair-trial protections while imposing discipline and sequencing on high-stakes proprietary technology disputes.