California's Algorithm Considered ZIP Codes in Vaccine Distribution, Allegedly Excluding Low-Income Neighborhoods and Communities of Color

California's vaccine-distribution algorithm used ZIP codes as opposed to census tracts in its decision-making, which critics said undermined equity and access for vulnerable communities who are largely low-income, underserved neighborhoods with low Healthy Places Index scores.

Risk classification

• Primary risk domain: 1 Discrimination & Toxicity
• Primary risk subdomain: 1.3 Unequal performance across groups

The vaccine distribution algorithm's effectiveness in prioritizing high-need areas was significantly lower for communities of color, resulting in unequal outcomes and reduced benefits for Latinx and Black populations.

Additional risk subdomains

• 7.3 Lack of capability or robustness: The algorithm lacked the geographic resolution (robustness) to accurately identify vulnerable pockets within larger ZIP codes, leading to system failure in critical resource allocation.

Causal factors

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

The inequitable distribution risk was caused by human design decisions to use ZIP codes instead of census tracts in the algorithm, which unintentionally undermined the state's equity goals during post-deployment planning.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The system is used in healthcare and critical public resource allocation (vaccine distribution), which has significant implications for safety, fundamental rights, and public interests.

AI system and alleged parties

• AI system: California vaccine-distribution algorithm
• AI purpose: Resource Allocation; Data Grouping
• Behaviour type: Tool
• Alleged developer: Blue Shield of California
• Alleged deployer: California Department of Public Health
• Alleged harmed parties: California low-income neighborhoods, California communities of color

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Substantial. 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 Substantial, indirect Substantial
• Civil rights: direct Negligible, indirect Severe
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Differential treatment

Reported: Yes, the report explicitly describes how the algorithm's design would result in unequal vaccine allocation, disproportionately omitting Latinx and Black communities from prioritization.

Directly caused: The algorithm directly prioritized larger ZIP codes over smaller census tracts, which would exclude over 2 million people in low-income neighborhoods from receiving additional vaccine supply.

Indirectly caused: This algorithmic exclusion disproportionately impacts communities of color, as Latinx people make up 53% and Black people make up 8% of the population in the omitted census tracts.

Civil rights

Reported: Yes, the report describes a potential violation of civil rights and equity principles in public health resource distribution.

Indirectly caused: The algorithm's exclusion of vulnerable communities of color undermines social justice and equitable access to life-saving medical care for over 2 million people.

People affected

• Occurrences reported: 1
• People reportedly exposed: 2000000

Potential causes

Management

• Operational Bias in Design: Blue Shield and state management prioritized deployment speed over equity.
• Inadequate Oversight of Vendor: Lack of rigorous state oversight on Blue Shield's algorithmic design choices.

Technology

• ZIP Code Resolution Flaw: Using ZIP codes instead of finer-grained census tracts diluted equity scores.
• Inflexible Allocation Logic: The system could not handle sub-ZIP code level targeting for vaccine supply.

Data Inputs

• Coarse Geographic Data: ZIP codes were used instead of socioeconomic census tracts for tracking.
• Omission of Demographic Factors: The algorithm did not explicitly prioritize high-risk racial or ethnic groups.

Human Factors

• Prioritizing Simplicity: Decision-makers favored operational ease over precise equitable distribution.

Process and Methods

• Inadequate Impact Assessment: The state failed to model how ZIP code grouping would impact vulnerable areas.
• Flawed Metric Application: Applying the Healthy Places Index to ZIP codes instead of census tracts.

Regulatory Environment

• Lack of Algorithmic Standards: No regulatory mandates existed to audit public policy algorithms for equity.

Information quality

• Classification confidence: High
• Reason for confidence: The report provides a clear, detailed analysis of the algorithm's design choices, the specific geographic metrics used, and the quantified demographic impact on the affected population. There is little ambiguity about the system's role or the nature of the risk.
• Ambiguities identified: The report analyzes a planned rollout, so the actual real-world epidemiological outcomes (exact infections/deaths resulting from the algorithm's deployment) are not documented.
• Alternative interpretations: The incident could be viewed purely as a policy/administrative failure rather than an AI safety failure, though the use of an automated algorithm to execute this policy makes it highly relevant to algorithmic safety.

An analysis of California's COVID-19 vaccine distribution algorithm developed by Blue Shield showed that using ZIP codes instead of census tracts risked excluding 2 million high-need individuals, disproportionately affecting communities of color. While representing a significant domestic policy and civil rights concern during a public health crisis, the incident does not present a direct threat to US national security.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Near-term. Occurred during an active public health crisis, requiring timely policy adjustment to prevent inequitable health outcomes.
• Autonomy: Human-controlled. The algorithm operated as a resource-allocation tool under direct human design, deployment, and policy oversight.
• Novelty: Established threat. Algorithmic bias and geographic proxies (like ZIP codes) causing demographic disparities is a well-documented and established issue in automated decision-making.

Impact by dimension

• Physical security: Minor. Algorithmic design choice affected vaccine distribution equity within a US state. While public health is critical infrastructure, this was an administrative policy issue rather than a kinetic attack or critical system compromise.
• Information security: Negligible. No evidence of information warfare, intelligence compromise, or foreign influence operations associated with this incident.
• Sovereignty: Minor. The incident involved a state-level public health decision-making process, but it did not threaten state authority, constitutional processes, or federal sovereignty.
• Economic security: Negligible. No indications of economic warfare, strategic technology theft, or systemic financial instability arising from this algorithmic choice.
• Societal stability: Minor. The algorithm caused systematic, unintentional discrimination in resource allocation against 2 million people in vulnerable communities. However, this is managed through domestic civil rights and administrative channels rather than national security apparatus.