Where Are the Other 249 Faces? Inside ACTIC Facial Recognition Searches in Arizona
This article discusses publicly documented Arizona government records, federal biometric standards, court filings, procurement documents, forensic guidelines, and allegations contained in pending federal civil litigation. Allegations are expressly identified as such. Nothing in this article claims that facial recognition was used in a specific Yavapai County case unless supported by an official public record.
You wake up in an interrogation room. You wake up at a traffic stop on Montezuma Street. You wake up with your mugshot vector-quantized into a 512-dimensional floating-point array inside a server rack at 2102 West Encanto Boulevard in Phoenix.
When the police lean across the table and tell you, “We identified you from a photograph,” they want you to believe in the absolute, unquestionable magic of modern forensic science.
They want you to fold. They want you to sign a plea or make an admission.
There is only one question your defense attorney needs to ask:
How?
Because when the state’s answer involves automated biometric algorithms running through state and federal surveillance nodes, that single question shatters into a hundred technical demands.
Arizona’s Counter Terrorism Information Center (ACTIC), a joint operational hub between the Arizona Department of Public Safety (AZDPS), the Arizona Department of Homeland Security (AZDOHS), and the FBI houses a specialized Forensic Images Unit (FIU). Public records confirm the FIU routinely executes biometric queries across tens of millions of state and federal records.
Here is the truth the state won’t put in the police narrative: A facial-recognition query does not generate a single, infallible name.
It generates a mathematical candidate list.
Then a human being with all their inherent biases, pressures, and subjective assumptions steps in to make a guess.
That exact boundary where machine scoring ends and human subjective selection begins is where police narratives routinely hide the truth.
Consider one remarkable Arizona case. According to sworn allegations in a 2026 federal civil rights complaint describing an underlying AZDPS facial-recognition report, a 2016 biometric search returned:
- 200 possible Arizona/DPS candidates, and
- 50 possible FBI Next Generation Identification (NGI) candidates.
An AZDPS examiner reviewed those lists, selected Javier Lorenzano-Nunez as a possible investigative lead, and documented seven visual similarities.
That means the central trial question is not: “Did the machine find him?”
The real question—the one that makes every prosecutor in Yavapai County sweat—is:
Where are the other 249 faces?
Were they ranked higher than the defendant? What were their exact similarity scores? Did another candidate score a 98% match while the defendant scored a 62%? Did the examiner document visual dissimilarities? Did a second examiner independently verify the match? Are the raw candidate galleries still preserved, or were they quietly wiped during a system migration?
These aren’t science-fiction questions. They are constitutional evidence questions under 28 C.F.R. Part 23 and the Fourth Amendment. For anyone facing prosecution in Prescott, Prescott Valley, Chino Valley, Cottonwood, Camp Verde, or anywhere in Yavapai County, understanding the true origin of an identification can mean the difference between a state prison sentence and a total case dismissal.
This issue connects directly to our master guide on:
What Is Parallel Construction? When Police Hide the Real Source of an Investigation
This infographic illustrates the critical gap between an algorithmic candidate list and a human-selected investigative lead and why rejected candidates, confidence scores, rankings, and examiner worksheets are crucial evidence in criminal defense.
1. The ACTIC Biometric Infrastructure Is Massive and Active
This is not speculation. ACTIC has operated continuously since October 2004 as Arizona’s primary fusion center, maintaining an unclassified multi-agency suite alongside a classified FBI Joint Terrorism Task Force (JTTF) suite.
ACTIC’s Forensic Images Unit (FIU) provides centralized biometric search capabilities to federal, state, tribal, and local law enforcement. Official agency documentation confirms that the FIU maintains direct search access to:
- Approximately 15.7 million Arizona booking images;
- Approximately 30 million Arizona driver-license and ID images;
- Approximately 64.7 million FBI NGI biometric records;
- The Arizona Missing and Exploited Children repository;
- A specialized state tattoo repository containing roughly 1.4 million images;
- Specialized databases such as Spotlight for human-trafficking inquiries;
- And the Homeland Security Information Network (HSIN) Multistate Facial Recognition portal for regional interstate referrals when local searches yield no leads.
Furthermore, ACTIC operates a statewide Threat Liaison Officer (TLO) network spanning local police departments, fire departments, and military resources. Municipal agencies across Arizona—such as the Cottonwood Police Department publicly document participation in the TLO program.
This means an unknown photograph captured by a local officer or surveillance system can propagate through state databases, federal repositories, and multi-state fusion channels long before an arrest warrant is ever drafted.
Read the official agency operational description here:
Arizona Counter Terrorism Information Center – AZ DPS Forensic Images Unit
2. Algorithmic Rank vs. Human Selection: The 250-Candidate Paradox
To cross-examine a biometric identification, counsel must isolate the distinct links in the chain:
PROBE IMAGE → ALGORITHMIC CANDIDATE LIST → HUMAN EXAMINER SELECTION → INVESTIGATIVE LEAD
Modern facial-recognition applications perform what is mathematically classified as a one-to-many ($1:N$) search. An unknown image (the probe image) is mathematically parsed into a biometric template and compared against millions of gallery templates. The system outputs a candidate gallery ranked strictly by algorithmic vector similarity.
The algorithm does not state: “This is the suspect.”
It states: “These gallery images are mathematically closest to the probe vectors based on our current feature-weight settings.”
Federal guidelines from the FBI and the Facial Identification Scientific Working Group (FISWG) state explicitly that facial-recognition search results are investigative leads only, requiring independent human evaluation and corroborating evidence prior to any enforcement action.
See official federal guidance:
FBI – Next Generation Identification (NGI) System
3. The Arizona Precedent: State v. Javier Lorenzano-Nunez
The clearest public illustration of this biometric workflow appears in the record of State v. Javier Lorenzano-Nunez (Maricopa County Superior Court No. CR2020-002309-001 DT) and its subsequent federal civil rights action (Lorenzano-Nunez v. Roestenberg et al., No. 2:26-cv-04153-ROS-DMF).
According to sworn pleadings in the 2026 federal complaint:
- 1998 Unsolved Homicide: Phoenix Police investigated the murder of Sarah Carr. Eyewitnesses were shown photo lineups and identified an Arizona MVD driver’s license photograph belonging to a man named Gilbert Noel Sanchez Rosado.
- The 2007 Search (No Match): In November 2007, police submitted Gilbert’s photograph to the facial-recognition unit operating at ACTIC. That search returned no match.
- The 2016 Search (250 Candidates): In November 2016, Phoenix Police Detective Dominick Roestenberg (Badge 6773) requested AZDPS to run Gilbert’s MVD photograph through upgraded biometric software against state databases and the FBI NGI system.
- The Result: The state query returned 200 possible Arizona/DPS candidates, while the federal search returned 50 possible FBI NGI candidates.
- Human Selection: AZDPS Sergeant Daniel Heltemes reviewed the candidate galleries, selected Javier Lorenzano-Nunez as a possible lead, and noted seven visual similarities.
- Intelligence Research: AZDPS Specialist Steffani Skelton performed follow-up intelligence research on Javier, expressly documenting that he had no known ties to Arizona.
Review the federal civil complaint:
Lorenzano-Nunez Federal Civil Complaint (U.S. District Court)
4. Semantic Drift: How a 1-in-250 Lead Became an Asserted Identity
What happened next demonstrates how probabilistic biometric leads can undergo semantic drift, morphing from a low-confidence investigative suggestion into an unassailable assertion of fact inside official record systems.
According to allegations in the civil complaint, on September 24, 2020, a Maricopa County Attorney’s Office employee sent an internal email stating: “Actually, Javier is an alias in Karpel [the prosecution case management database]. His name in Karpel is Gilbert Rosado.”
When the case was presented to a grand jury in late 2020, the detective effectively substituted Javier’s identity for Gilbert’s, testifying as if witnesses had originally identified Javier himself.
On February 26, 2025, Maricopa County Superior Court Judge Aryeh D. Schwartz issued a formal minute entry granting a defense motion to remand for a new probable cause determination. The court held that the presentation of identity evidence to the grand jury was materially misleading and violated due process.
Read the Superior Court’s official order:
Maricopa County Superior Court – February 26, 2025 Remand Order
On August 5, 2025, the State moved to dismiss the criminal case entirely without prejudice. On June 11, 2026, Lorenzano-Nunez filed his federal civil rights lawsuit.
Read the official dismissal record:
Maricopa County Superior Court – August 5, 2025 Dismissal Minute Entry
The Forensic Takeaway: The court did not hold that facial recognition software is unconstitutional per se. Rather, the case highlights the massive due-process danger when police and prosecutors treat a human-selected biometric lead as a proven identity while ignoring contradictory intelligence and missing candidate galleries.
5. The Native FBI Architecture: Technical Specifications Disclose What Records Exist
When challenging an ACTIC biometric search, defense counsel should not accept a sanitized, one-page summary narrative. The federal government publishes detailed technical specifications defining exactly what machine transactions occur during a search.
Under the FBI Electronic Biometric Transmission Specification (EBTS v10.0.7), automated facial searches generate structured Electronic Biometric Transmission transactions:
- FRS (Facial Recognition Search Request): The native transmission containing the probe image, Originating Agency Identifier (ORI), Transaction Control Number (TCN), and parameter filters.
- SRB (Biometric Search Response): The structured return containing the Candidate Investigative List. Under federal specifications during the relevant period, an SRB could return a maximum candidate gallery of exactly 50 facial images alongside candidate User Control Numbers (UCN), match scores, and rankings.
- BDEC (Biometric Candidate Decision Feedback): A standardized transaction allowing local agencies to transmit candidate-disposition feedback back to federal systems.
Review the federal specifications:
FBI Electronic Biometric Transmission Specification (EBTS)
Furthermore, Arizona’s internal biometric architecture has evolved significantly over time. Procurement records show that ACTIC utilized Morpho Face Examiner around 2020 before transitioning to the cloud-native IDEMIA Arizona Biometric Information System (ABIS) under state project PS20003, which went live in June 2022.
If your case involves an older search, software versioning and system migration logs become critical targets for discovery.
6. Parallel Construction: How Fusion Center Leads Are Hidden
Facial recognition rarely appears on page one of an arrest report. Instead, it frequently operates as an invisible trigger for downstream surveillance.
Imagine this typical investigative chain:
Surveillance Still → ACTIC FIU Search → 250 Candidates → Analyst Selection → Address Lookup → Traffic Stop → Arrest
When the officer writes the incident report, the narrative begins at the traffic stop: “On October 12, officers observed a vehicle commit a lane violation…” The biometric query that initiated the entire chain disappears—a tactic known as parallel construction.
Learn how to expose hidden investigative origins:
What Is Parallel Construction? When Police Hide the Real Source
For another example of hidden database correlation, see:
Can Police Link Your Phone to Your Car? SignalTrace and Device Correlation Explained
7. The Master Defense Discovery Package: 25 Mandatory Demands
If facial recognition played any role in your case, defense counsel must demand the entire audit trail rather than accepting a simple summary report.
I. The Probe Image & Preprocessing
- The native, uncompressed original source image or video frame.
- Complete EXIF, file metadata, and cryptographic hashes (SHA-256) for all image iterations.
- Records of all image transformations: cropping, rotation, brightness/contrast adjustments, landmark placements, or pose-normalization filters.
II. Algorithmic Search Audit Logs
- Native FRS submission records and SRB response files.
- Transaction Control Numbers (TCN), Transaction Control References (TCR), and Agency ORIs.
- Software vendor name, client application version, facial-engine SDK build, and algorithm version.
- Exact search parameters: candidate limits, similarity threshold settings, and demographic filters.
- Logs of all search reruns, parameter modifications, or failed queries.
III. The Rejected Candidate Galleries
- The complete candidate gallery returned by each state, federal, or multistate query.
- Algorithmic similarity scores and rank orderings for every returned candidate.
- High-resolution photographs and biographic identifiers for all rejected candidates.
- Written examiner notes documenting why higher-ranked candidates were excluded.
IV. Human Examiner Worksheets & Methodology
- The primary examiner’s complete benchmark worksheet and feature-comparison notes.
- Documented visual similarities and all documented dissimilarities.
- Compliance documentation under FISWG Minimum Guidelines for Facial Image Comparison Documentation.
- Evidence of whether candidate scores or biographic names were visible to the examiner during visual review (contextual bias logs).
- Independent second-examiner review records and blind verification logs.
- Examiner proficiency testing, error rates, and vendor certification records.
V. Intelligence Research & Case Management Systems
- All follow-up intelligence research logs, analyst notes, and query histories (e.g., ACTIC Specialist research).
- Contradictory intelligence findings (e.g., documented lack of geographic ties).
- Prosecutor case management system audit trails (e.g., Karpel audit logs tracking alias creation and identity modifications).
- Interagency communications, emails, and P3 Tips submission logs.
- System retention schedules, purge logs, migration reports (e.g., 2022 ABIS cloud migration), and legacy archive inventories.
- Compliance records under 28 C.F.R. Part 23 governing reasonable suspicion and criminal intelligence retention.
- Complete custodian declarations verifying system searches across active, archived, and backup repositories.
8. Frequently Asked Questions About ACTIC Facial Recognition
Does ACTIC actively perform facial-recognition searches?
Yes. ACTIC publicly confirms that its Forensic Images Unit (FIU) conducts facial and tattoo recognition for law enforcement, searching state booking repositories, driver’s license records, FBI NGI databases, and regional fusion networks.
Is a facial-recognition candidate match considered a positive identification?
No. Facial-recognition software executes one-to-many searches that return ranked candidate lists based on mathematical vector similarity. Federal agencies and forensic standards organizations emphasize that candidate matches are investigative leads only, requiring independent visual examination and corroborating physical evidence.
Why do rejected candidates matter in an Arizona criminal case?
If an algorithm generates 250 possible candidates, the rejected candidates form the baseline for evaluating human selection. If a candidate ranked #1 or #5 possessed a higher similarity score, shared additional facial characteristics, or matched suspect descriptors, that evidence may be highly exculpatory under Brady v. Maryland.
Did a court rule that Arizona’s facial-recognition software failed in the Lorenzano-Nunez case?
No. On February 26, 2025, the Maricopa County Superior Court granted a remand because police and prosecutors presented identity evidence to the grand jury in a materially misleading manner by substituting identities. The court did not rule on the underlying algorithm’s technical accuracy.
How can a criminal defense attorney challenge facial-recognition evidence in Yavapai County?
Defense counsel can file targeted motions for discovery under Rule 15, demanding raw transaction logs (FRS/SRB), algorithm versions, candidate galleries, similarity scores, examiner worksheets, dissimilarity notes, and verification records. If the state failed to preserve candidate galleries or concealed the search origin, counsel may move for suppression or dismissal based on due process and spoliation of evidence.
Facing Charges in Yavapai County? Audit the Identification.
A police narrative stating that investigators “developed a lead” is the beginning of a legal defense and not the end.
Whether your case originated in Prescott, Prescott Valley, Chino Valley, Cottonwood, Camp Verde, or anywhere across Yavapai County, you have the right to inspect the machine logic, candidate scores, examiner notes, and rejected faces that police relied on.
Start your legal strategy here:
Prescott Criminal Defense Lawyer – Ted Agnick
Understand the procedural roadmap:
Criminal Case Stages in Prescott, AZ
Call 928-776-1782
Ted Agnick | DUI & Criminal Attorney
140 N Montezuma Street
Prescott, AZ 86301
