{"resourceId":"washington-county-or-language-access-framework-ai-limits-20260930","versions":[{"version":"external-2fd3d1aaec7e4f32578706b3461f7781f8d7eca6031a0aeb234cecc55329b0a1","resource":{"id":"washington-county-or-language-access-framework-ai-limits-20260930","title":"Washington County, Oregon's 2026 Language Access Framework limits AI translation to brief contacts, requires human review, disclosure and a no-training clause","organization":"Washington County, Oregon, Office of Access and Opportunity","sector":"County language access policy","geography":"Washington County, Oregon, United States","publishedAt":"2026-09-30 (date of certification)","publicationDate":"2026-09-30","eventDate":"2026-09-30","sourceName":"Washington County, Oregon","sourceLabel":"Washington County Language Access Framework 2026 (PDF)","sourceUrl":"https://washingtoncountyda.org/index%2ephp/oao/documents/language-access-framework/download?inline=","evidenceClass":"standards-guidance","outcomeClass":"emerging","topics":["governance-procurement","accessibility-workforce","data-security"],"finding":"The framework, certified Sept. 30, 2026 for 2026 to 2028, allows AI and automated translation only for brief initial point-of-contact exchanges and says these tools are not an acceptable substitute for qualified interpretation in substantive interactions. Machine translation may not be the sole method for rights and benefits information, eligibility, deadlines, legal notices, emergency instructions, complaint intake or public participation materials. Residents must be told at the outset when an AI tool is used and offered a qualified interpreter. Departments must notify the county before piloting AI language tools and log translation errors. The county evaluates vendor data practices and requires a contractual ban on using county interactions for model training unless an exception is granted. Vital webpages must be professionally translated.","sledRelevance":"A concrete, adoptable county policy for AI in language access, newly relevant this week as California's courts cancelled an AI counter-translation rollout. It gives other counties and cities model language for disclosure, human review, error logging and vendor data terms.","evidence":"Primary policy document; describes requirements, not measured outcomes. Implementation timelines run from adoption over 30 to 180 days.","architectureImplications":"Website machine translation stays available for general understanding, but vital pages and forms need professional translation; AI tools must support disclosure and handoff to human interpreters.","governanceImplications":"Central review of any AI language pilot, error logging, and vendor data-governance review before approval.","securityPrivacyImplications":"Requires review of where conversation data is stored, whether it trains models and whether it leaves the United States, with a default contractual ban on training use.","caveats":"Policy, not evaluation; compliance and enforcement not yet observable; written for one county's Title VI obligations.","streamIds":["citizen-engagement"],"roles":{"sales":"This gives buyers a policy model, not a product case. A county has written clear limits on AI translation: brief contacts only, disclosure, human review and no vendor model training on county data. For SLED clients, this frames discovery around which interactions qualify and what contract terms are required. A bounded engagement is helping a jurisdiction adapt similar policy and vendor terms. It shows no measured benefit.","engineering":"Engineering implications follow from the rules: tools must disclose AI use at the start of an interaction in the resident's language, hand off to a human interpreter on trigger topics, log reported errors, and keep data in approved locations without training use. Website translation widgets stay acceptable for general content but not for vital pages. Verify vendor data flows and contract terms before any pilot.","delivery":"Delivery work is policy rollout and monitoring: a vital document and vital webpage inventory, scripts for front-desk staff on trigger topics, an approval process for AI language pilots, an error log reviewed by the access office, and vendor contract updates. Acceptance should measure compliance with disclosure and handoff, not just translation volume. These are proposals; the framework reports no results yet."},"roleDetails":{"sales":{"customerProblem":"Agencies need clear rules for when AI translation is acceptable for residents with limited English.","stakeholders":["Language access or equity officer","Department directors","County counsel","Procurement and IT"],"discoveryQuestions":["Do you have written limits on AI translation for rights, benefits and safety topics?","Do vendor contracts prohibit training on resident interactions?","How are AI translation errors reported and reviewed?"],"valueHypothesis":"Clear policy and contract terms could lower legal and trust risk when adopting AI language tools; benefit is not measured.","potentialEngagement":"A policy and contract-terms review that adapts this framework's AI provisions to the client's jurisdiction.","unsupportedClaims":["That AI translation is safe for vital documents","Any outcome data from Washington County","Applicability without legal review in other states"]},"engineering":{"fit":"Applies to AI interpretation, chat translation and website translation used by any county department.","architectureIntegration":"Require disclosure at start, trigger-topic handoff to human interpreters, error logging and data-location controls in each tool.","prerequisites":["Vital document and webpage inventory","Interpreter service on call","Vendor data-flow documentation"],"constraints":["Title VI obligations","Safe Harbor language list","Vendor platforms may route data outside the United States"],"security":"Contractual ban on model training with county interactions and review of storage location and data transfer.","proofOfValue":["Confirm disclosure appears in each supported language","Test that trigger topics hand off to a human","Audit vendor data handling against the contract"]},"delivery":{"work":"Roll out AI language-tool governance across departments.","dependencies":["Access office staff","County counsel","Vendors and interpreter services"],"ownership":"County Office of Access and Opportunity, with department language liaisons.","skillsAdoption":"Front-desk staff need training on trigger topics and disclosure scripts.","governanceCheckpoints":["Pre-pilot notification and approval","Vendor data-governance review","Periodic error-log review"],"acceptanceCriteria":["Proposed: 100% of approved AI tools disclose AI use at the start of each interaction","Proposed: vital webpages professionally translated on the framework timeline"],"risks":["Staff using unapproved tools","Errors not reported","Vendor terms not updated"]}},"retrievedAt":"2026-10-31T03:05:09Z","enrichedAt":"2026-10-31T03:08:32Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Front-desk staff must recognize topics that trigger an immediate move to a qualified interpreter; the framework also covers digital accessibility standards.","procurementImplications":"Vendors may not use AI machine translation for county services without prior written approval and disclosure.","operatingModelImplications":"Puts the county access office in charge of approving, monitoring and evaluating AI language tools across departments.","sourceVerification":{"openedUrl":"https://washingtoncountyda.org/index%2ephp/oao/documents/language-access-framework/download?inline=","referenceExcerpt":"Departments must ensure human review when machine translation is used for public-facing","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}