Portland Water Bureau SERVUS Algorithm Reportedly Allocates Utility Bill Discount to High-Wealth Consumer

The Portland Water Bureau's AI-driven pilot program for water bill discounts is reported to have randomly selected Tim Boyle, a wealthy high-water consumer, for a 40% discount intended for financially struggling customers. The program, developed by SERVUS, is meant to identify underserved individuals by using machine learning.

Risk classification

• Primary risk domain: 7 AI system safety, failures, & limitations
• Primary risk subdomain: 7.3 Lack of capability or robustness

The AI system failed to perform reliably by selecting an extremely wealthy individual for a low-income assistance program, demonstrating a lack of capability in accurately targeting the intended demographic.

Causal factors

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

The selection of a wealthy CEO for a low-income discount was an unexpected and unintentional outcome of the deployed SERVUS algorithm.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The AI system is used in critical infrastructure and the distribution of essential public services (water utility assistance).

AI system and alleged parties

• AI system: SERVUS algorithm (SERVUS)
• AI purpose: Resource Allocation; Identification
• Behaviour type: Assistant
• Alleged developer: Portland Water Bureau
• Alleged deployer: Portland Water Bureau
• Alleged harmed parties: Tim Boyle, Portland Water Bureau, Low-income Portland residents, City of Portland

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 Negligible
• Civil rights: direct Negligible, indirect Negligible
• 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: The report describes the system selecting a wealthy individual for a discount intended for low-income customers, representing an unintended differential treatment in resource allocation.

Directly caused: Tim Boyle was selected for a 40 percent water bill discount while actually needy customers may have been missed during this pilot phase.

People affected

• Occurrences reported: 1
• People reportedly exposed: 1

Potential causes

Management

• Inadequate Risk Assessment: Management failed to foresee the PR risk of offering discounts to billionaires.
• Poor Program Oversight: Approval of a randomized trial without safety boundaries for high-net-worth.

Technology

• Unconstrained Randomization: The software randomly selected accounts without filtering out wealthy outliers.
• Lack of Algorithmic Filters: The ML pilot lacked filters to exclude high-volume water consumers.

Data Inputs

• Unfiltered Customer Input Pool: The dataset included all customers regardless of income or water usage.
• Missing Financial Data: The system did not utilize financial indicators to pre-screen candidates.

Human Factors

• Overreliance on Automation: Staff trusted the automated random selection without manual verification.

Process and Methods

• Flawed Pilot Trial Design: The randomized control trial design included extremely wealthy individuals.
• No Pre-Screening Process: The bureau did not screen the recipient list before mailing discount offers.

Information quality

• Classification confidence: High
• Reason for confidence: The report clearly outlines the pilot program, the role of the SERVUS algorithm, and the specific outcome involving Tim Boyle. There is little ambiguity about the facts of the incident.

An AI-driven pilot program by the Portland Water Bureau mistakenly allocated a low-income water bill discount to a wealthy CEO. The incident is a localized administrative and algorithmic failure with negligible implications for national security.

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

Threat characteristics

• Imminence: Long-term. The incident represents a localized, resolved administrative error with no active or impending national security threat.
• Autonomy: Human-supervised. The AI system autonomously selected recipients and generated letters, but operated within a pilot framework overseen by municipal authorities.
• Novelty: Established threat. Algorithmic errors in public benefit distribution and demographic targeting are well-documented and established issues.

Impact by dimension

• Physical security: Negligible. The incident involved a water bill discount pilot program and did not impact the physical operations or safety of critical water infrastructure.
• Information security: Negligible. No intelligence compromise, classified data theft, or information warfare operations were associated with this billing algorithm error.
• Sovereignty: Negligible. This was a local municipal administrative error in Portland, Oregon, with no impact on national sovereignty or core federal government functions.
• Economic security: Negligible. The financial impact was limited to a local pilot program contract of 350,000 dollars, presenting no threat to national economic or technological security.
• Societal stability: Negligible. While the algorithm misallocated a discount intended for low-income residents, it did not cause large-scale societal instability or systemic human rights abuses.