Canadian Police's Release of Suspect's AI-Generated Facial Photo Reportedly Reinforced Racial Profiling

The Edmonton Police Service (EPS) in Canada released a facial image of a Black male suspect generated by an algorithm using DNA phenotyping, which was denounced by the local community as racial profiling.

Risk classification

• Primary risk domain: 1 Discrimination & Toxicity
• Primary risk subdomain: 1.1 Unfair discrimination and misrepresentation

The AI system generated a highly generic representation of a Black male suspect based on default parameters, which unfairly represented and stereotyped the local Black community, exposing them to broad suspicion.

Additional risk subdomains

• 7.3 Lack of capability or robustness: The DNA phenotyping technology lacks the scientific capability to accurately reconstruct unique human faces, age, or BMI, leading to inaccurate and speculative outputs.
• 5.1 Overreliance and unsafe use: Law enforcement relied on an unvalidated, proprietary AI tool in a critical criminal investigation, ignoring the scientific limitations of the technology.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The incident was triggered by the output of the Snapshot AI system, which generated a highly generic and potentially biased facial composite, leading to unintended public backlash and community harm.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The AI system is used in a law enforcement application for biometric prediction and suspect profiling, which has significant implications for fundamental rights and public safety.

AI system and alleged parties

• AI system: Snapshot DNA Phenotyping Service (Parabon NanoLabs)
• AI purpose: Identification; Image Generation
• Behaviour type: Tool
• Alleged developer: Parabon Nanolabs
• Alleged deployer: Edmonton Police Service
• Alleged harmed parties: Black residents in Edmonton

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Minor. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Minor, indirect Minor
• Civil rights: direct Negligible, indirect Minor
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Minor
• Psychological: direct Negligible, indirect Minor
• Epistemic: direct Minor, indirect Minor
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Differential treatment

Reported: The report explicitly describes concerns regarding racial profiling and the potential for the visual profile to provide too broad a characterization of Edmonton's Black community.

Directly caused: The AI generated a generic Black male composite using default parameters, which disproportionately targeted a specific racial group.

Indirectly caused: The publication of the image by police subjected the local Black community to systemic bias and potential discriminatory scrutiny.

Inferred additional harm: The incident likely increased the risk of differential treatment and wrongful stops of Black individuals matching the generic description by law enforcement or the public.

Civil rights

Reported: The report explicitly describes concerns about profiling of a marginalized group and the violation of civil rights.

Indirectly caused: The dissemination of an unscientific, generic suspect image threatened the civil rights of Edmonton's Black community by exposing them to arbitrary suspicion.

Inferred additional harm: The use of unvalidated biometric prediction tools in law enforcement poses a long-term threat to civil liberties and the right to be free from arbitrary discrimination.

Privacy

Reported: The report explicitly describes privacy concerns regarding the use of DNA databases and the potential exposure of user data without consent.

Indirectly caused: The broader practice of forensic genetic genealogy described in the reports raised significant privacy concerns among database users and genealogists.

Psychological

Reported: The report explicitly describes public backlash, uproar, and concern within the marginalized Black community of Edmonton.

Indirectly caused: The release of the generic suspect image caused distress, anxiety, and feelings of being racially profiled among members of Edmonton's Black community.

Inferred additional harm: It is likely that many Black residents in Edmonton experienced heightened anxiety and fear of wrongful police contact due to the dissemination of the generic composite.

Epistemic

Reported: The report explicitly describes the dissemination of an unscientific and potentially inaccurate facial composite as a representation of a real suspect.

Directly caused: The AI generated a highly speculative facial composite based on unvalidated phenotyping models.

Indirectly caused: The police published this speculative image, potentially misleading the public and creating false beliefs about the suspect's actual appearance.

People affected

• Occurrences reported: 1
• People reportedly exposed: 22000

Potential causes

Management

• Prioritizing Lead Over Harm: Management accepted the trade-off of profiling the Black community.
• Inadequate Risk Assessment: COO admitted they did not properly consider the risks to the community.

Technology

• Unvalidated Phenotyping AI: The AI's ability to predict faces from DNA is unvalidated in literature.
• Inability to Determine Age/BMI: Using default values created an inaccurate and misleading portrait.
• AI Generates Generic Faces: The model tweaks generic faces rather than predicting unique features.

Data Inputs

• Degraded DNA Samples: Scant DNA forces the model to anticipate blank spots with algorithms.
• Lack of Non-Genetic Data: DNA cannot predict non-genetic factors like smoking, diet, or age.

Human Factors

• Overreliance on AI Output: EPS published the composite as a last resort without realizing the risks.
• Lack of Public Risk Awareness: The department did not foresee the backlash regarding racial profiling.

Process and Methods

• Lack of Scientific Peer Review: The lack of open validation makes the technology's accuracy questionable.
• Insufficient Internal Screening: EPS lacked tools to stress-test decisions before releasing AI portraits.

Regulatory Environment

• Lack of Phenotyping Regulations: Absence of regulatory oversight allows the use of unvalidated AI tools.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide consistent and detailed coverage of the Edmonton Police Service's use of the Snapshot tool, the public backlash, and the subsequent apology. The scientific criticisms of DNA phenotyping are clearly articulated by experts in the text.
• Ambiguities identified: The exact number of people in the Edmonton Black community who felt impacted or distressed is not quantified.

The Edmonton Police Service used an AI-based DNA phenotyping tool to generate and release a generic suspect composite, leading to public backlash and accusations of racial profiling. The incident highlights minor national security risks related to societal stability and civil rights, specifically the potential for AI-driven biometrics to reinforce systemic bias when deployed by law enforcement.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: Canada
• Alleged perpetrator: Edmonton Police Service

Threat characteristics

• Imminence: Long-term. The immediate incident has been resolved with an apology and retraction, but the strategic policy concerns regarding AI biometrics in policing remain.
• Autonomy: Human-controlled. The AI tool generated a static composite based on DNA inputs, but human law enforcement officers made the final decision to publish and distribute the image.
• Novelty: Evolved capability. Represents an evolutionary advancement in forensic investigation methods by using algorithmic DNA phenotyping to generate suspect composites, replacing traditional sketch artists.

Impact by dimension

• Physical security: Negligible. No physical security threats, kinetic attacks, or critical infrastructure disruptions were associated with this incident.
• Information security: Negligible. The incident did not involve intelligence compromise, espionage, or state-sponsored information operations.
• Sovereignty: Negligible. Core government operations and sovereign authority remained unaffected, as this was a localized municipal law enforcement investigative decision.
• Economic security: Negligible. No strategic technology theft, economic warfare, or critical supply chain vulnerabilities were identified.
• Societal stability: Minor. The release of the AI-generated composite sparked localized public backlash and raised significant civil rights and racial profiling concerns within the Edmonton Black community, but did not threaten national social cohesion.