From the State Government edition of October 26, 2026
Washington's secretary of state wants to pilot AI ballot signature verification as early as 2027, with humans reviewing every AI rejection and spot-checking approvals
Washington Office of the Secretary of State · State election administration; mail-ballot signature verification · Washington, U.S.
- Publisher
- Government Technology
- Original publication
- 2026-10-26
- Source retrieved
- 2026-10-27
What happened
Secretary of State Steve Hobbs wants to test AI from the financial industry to compare mail-ballot envelope signatures with voter registration records, possibly as a 2027 pilot. Washington's 39 counties now rely on trained election workers, and a University of Washington study found higher challenge rates among Black, Asian, Hispanic and younger voters. Hobbs proposes human review of every signature the AI rejects and random checks of signatures it approves. The state elections director said planning is preliminary, with talks held with multiple vendors and counties, and work resumes in December or early 2027.
Why it matters
Why now: reported Oct 26 and not in our archive. It is a state agency proposing AI for a high-stakes, rights-affecting determination with an explicit oversight design, which is the open question from our Oct 25 edition about what the human in the loop actually does. It also tests whether AI can reduce known disparities in manual review or would reproduce them.
Evidence and measured results
One news report opened in full, based on a broadcast interview and an agency statement. No vendor named, no accuracy data, no pilot design document. The University of Washington disparity study is cited but was not opened.
Limitations and uncertainty
Proposal stage only; no pilot approved, no vendor selected, no accuracy or bias testing reported. Relies on a single news report of an interview.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-26; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Manual signature verification produces uneven challenge rates across voter groups, and the state wants to test whether AI can help without losing human control over rejections.
- Stakeholders
- Secretary of State and state elections director
- County auditors and election staff
- Voting rights advocates
- State CIO and security office
- Discovery questions
- What accuracy and disparity thresholds would justify a pilot?
- Which counties would participate and with what ballot volumes?
- How large would the random sample of AI approvals be?
- Where would signature images be processed and stored?
- Value hypothesis
- Interpretation: AI-assisted matching with full human review of rejections might reduce reviewer workload and inconsistency, but this is untested in Washington and could also reproduce disparities.
- Potential engagement
- Bounded pilot design support: test plan, demographic error analysis method, sampling design for approvals and rejection review workflow, ahead of any vendor selection.
- Unsupported claims
- No evidence AI reduces challenge disparities
- No accuracy data for the financial-industry tools in elections
- No vendor or cost information
Pre-sales engineering
Role takeaway
- Fit
- Mail-ballot signature comparison against registration records in a state with county-run elections.
- Architecture and integration
- Score envelope signatures against registration images; route all rejections to human reviewers; randomly sample approvals for human audit; log scores, decisions and reviewer outcomes for error analysis.
- Prerequisites
- Representative labeled signature test set
- County data-sharing agreements
- Defined sampling rate for approvals
- Security review of vendor hosting
- Constraints
- Counties run elections, not the state
- Valid unconventional signatures
- Election-period freeze on changes; work resumes December or early 2027
- High public scrutiny
- Security
- Treat signature images and voter records as sensitive election data; require hosting, access logging and retention terms before testing.
- Proof of value
- Measure false rejection rates by voter group against manual review on historical ballots
- Test valid unconventional signatures in the evaluation set
- Estimate reviewer workload under full rejection review plus approval sampling
Delivery
Role takeaway
- Work
- Plan a controlled pilot with participating counties: data agreements, test protocol, review workflow, sampling, voter cure communications and evaluation reporting.
- Dependencies
- County participation
- Vendor willingness to support bias testing
- Legal review of election rules
- Post-election scheduling
- Ownership
- Washington Office of the Secretary of State Elections Division, with county auditors running ballot processing.
- Skills and adoption
- Election staff need training on reviewing AI rejections and sampled approvals without deferring to the score.
- Governance checkpoints
- Accuracy and disparity thresholds set before vendor testing
- Human review of every AI rejection
- Random audit of AI approvals
- Public reporting of pilot results
- Acceptance criteria
- Proposed: false rejection rate by voter group no worse than manual review on the test set
- Proposed: 100% of AI rejections reviewed by a person before voter notice
- Proposed: approval sampling rate and results published
- Risks
- AI reproduces or worsens demographic disparities
- Reviewers defer to AI scores
- Cybersecurity or vendor access issues
- Public trust damage if errors surface during an election
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
The proposal would sit between county ballot processing and the voter registration signature store, flagging mismatches for human review. It must handle legitimate unconventional signatures, such as a voter who signs with a drawing that matches his driver's license.
Governance
Who approves, reviews and stays accountable for outcomes?
Proposed oversight: humans review all AI rejections and randomly sample AI approvals. Counties administer elections, so a state pilot needs county participation; the agency says it has talked with multiple counties.
Security and privacy
What data, permissions and controls need testing?
Hobbs acknowledged cybersecurity concerns. Signature images and voter records are sensitive election data; vendor hosting and access terms are not described.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Election workers currently review signatures; the proposal shifts them toward reviewing AI rejections and sampled approvals.
Procurement
What should contracts, pricing and exit terms secure?
The agency has talked with multiple vendors; no procurement has started. Interpretation: bias testing and data hosting terms should be in any solicitation.
Operating model
Which teams own the service once it runs?
A state-led tool used inside county-run election processes requires shared ownership between the Secretary of State and county auditors.
Publication history
- 2026-10-26State Government · Issue 493 resources
Stable resource ID: washington-sos-proposes-ai-ballot-signature-verification-pilot-2027-20261026