State & Local GovernmentIssue 184 records
State Government
Four sources pair same-week state frontier-AI procurement and chatbot-governance moves with measured benefits-delivery chatbot volume. Oregon EO 26-26 (2026-09-23) directs the State CIO to propose, within 90 days, third-party AI safety-review standards and a kill-switch viability assessment for frontier models purchased by state government. Maryland’s Sept. 22 Moore AI framework tasks the AI Subcabinet with frontier-company legislative recommendations including independent audits and 72-hour incident reporting. California EDD’s July 24 Benefiting Californians update reports >99,000 unemployment chatbot interactions since May 2026 with nearly 34,000 fully resolved via self-service. FPF’s September 2026 chatbot-legislation landscape finds 37 states introduced 124 chatbot bills in 2026 (18 enacted across 2025–26), shifting from disclosures toward crisis protocols and minor/relational safeguards.
- Evidence records
- 4
- Cross-source patterns
- 1
- Evidence classes
- 3 government evaluation1 independent research
- Outcomes
- 3 emerging1 mixed
- Source freshness
- 2 new this fortnight2 recent
- Research completed
- 2026-09-24
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Frontier procurement controls are rising beside measured benefits chatbots and a dense chatbot-statute wave
Oregon EO 26-26 and Maryland’s Moore framework push third-party frontier AI reviews, kill-switch study, audits, and 72-hour incident reporting into state agendas, while California EDD already reports six-figure unemployment chatbot volume with tens of thousands of self-service resolutions, and FPF documents 124 chatbot bills in 2026 shifting toward interaction controls. Treat citizen-service scale metrics and frontier purchase gates as coupled workstreams—not substitutes.
Operating questionWhen a state scales a benefits or resident chatbot, what parallel frontier-procurement evaluation and chatbot-statute carveout checks are required before the next expansion?
Supporting evidenceOffice of the Governor of Oregon (Tina Kotek) / State Chief Information OfficerOffice of Governor Wes Moore (Maryland)California Employment Development Department (EDD)Future of Privacy Forum (FPF)
Evidence ledger
Full record · every source keeps its link and limitations
Oregon EO 26-26 directs third-party frontier AI safety reviews and kill-switch assessment for state procurement
Why now: on 2026-09-23 Governor Kotek signed EO 26-26, stating Oregon policy to seek frontier AI models that have received independent third-party AI safety review. The State CIO must, within 90 days, submit an implementation proposal covering a frontier-model definition; standards for adequate third-party safety review; evaluation of a kill-switch requirement’s viability; methods to assess reviewer independence/qualifications/methodology/transparency; integration into procurement and technology-use practices; exceptions when review is unavailable; ongoing monitoring; transparency recommendations; and any needed statutory/administrative/budgetary changes. Order is effective immediately, remains until terminated, and must be reassessed every three months. Builds on EO 23-26 AI Advisory Council work and existing Responsible AI Usage Policy.
Why it matters, evidence and limitations
- Why it matters
- Direct state-government procurement control for frontier models—transferable to other states designing purchase gates for advanced GenAI without waiting for federal standards; explicitly uses contracting power rather than only agency-use policy.
- Evidence and measured results
- Primary: four-page official EO PDF OCR’d after HTTPS download. Corroborated by KVAL (updated 2026-09-23) and EIN/Oregon DEQ release summarizing the same directives. No measured pilot outcomes in the order itself.
- Limitations and uncertainty
- Standards and kill-switch requirement are not yet operational—only a 90-day proposal mandate. OCR of scanned PDF may contain minor transcription artifacts; prefer final oregon.gov posting when available. Political statements about federal leadership are not treated as empirical findings.
Maryland Moore AI framework seeks frontier audits, 72-hour incident reporting, and worker voice
Why now: on 2026-09-22 Governor Moore announced an AI framework organized around protecting people, centering workers, and keeping kids safe. Under protecting people, the AI Subcabinet will develop legislative recommendations for regulating frontier AI model companies—including safety frameworks and testing, worker protections, independent third-party audits and evaluations, whistleblower protections, public reporting, and 72-hour incident reporting. Additional actions span Right of Publicity, civil-rights guidance for AI in housing/lending/employment, algorithmic tenant-screening limits on DHCD-financed properties, anti-rent-collusion legislation, gambling targeting limits, a standing Labor-Management Committee on AI for state workers, and stronger protections related to unsafe chatbots for youth.
Why it matters, evidence and limitations
- Why it matters
- State-level frontier-developer oversight agenda that mirrors elements other states are operationalizing (audits, incident reporting) while adding explicit state-workforce labor engagement—useful for multi-state policy and shared-services planning with clear 'not yet law' limits.
- Evidence and measured results
- Official press release dated September 22, 2026 (updated same day). Builds on 2024 EO, AI Governance Act of 2024, and Responsible AI Policy. No quantitative agency AI outcome metrics in the release.
- Limitations and uncertainty
- Most items are planned legislative recommendations or agency guidance tasks, not enacted requirements as of the release. Housing/gambling/chatbot provisions need future bill text. No measured results.
California EDD reports 99,000+ unemployment chatbot uses with nearly 34,000 self-service resolutions
Why now: although published 2026-07-24 (before the last edition), EDD’s measured unemployment chatbot volume was not previously archived and is newly relevant alongside 2026-09-22/23 state frontier-procurement and chatbot-governance moves—showing benefits-delivery self-service AI already scaling while frontier purchase controls tighten. Since May 2026, customers used the chatbot more than 99,000 times for claim-specific information; nearly 34,000 interactions were fully resolved through self-service without a live agent. Chatbot scopes claim status, payment details, and eligibility for claims filed within three years, with guided follow-ups. Separately, April 2026 certification plain-language updates cut Help-icon clicks 54.5% in three months—not attributed here as a chatbot KPI. August 2026 modernization PDF repeats 99,000+ chatbot uses without restating the 34,000 figure.
Why it matters, evidence and limitations
- Why it matters
- Concrete state benefits chatbot operating metrics for high-stakes unemployment claims—transferable to other state UI/contact-center modernizations, with clear limits that resolution≠accuracy and eligibility answers still need human/legal oversight design.
- Evidence and measured results
- Agency blog metrics; August PDF corroborates 99,000+ usage. No independent audit of answer accuracy, fraud resistance, or demographic disparity. Live-agent deflection implied by self-service resolution wording but not labeled as a percentage in source.
- Limitations and uncertainty
- updateExplanation: first-time archive of July metrics because newly salient against same-week frontier procurement EOs; not a same-week publication. Metrics are self-reported; 34,000 resolutions lack published accuracy sampling. Do not conflate Help-icon 54.5% drop with chatbot performance.
FPF finds 37 states introduced 124 chatbot bills in 2026 as laws shift from disclosures to interaction controls
Why now: FPF’s September 2026 report (methodology: bills tracked through September 2026) quantifies an active state chatbot-law wave—37 states introduced 124 chatbot-related bills in 2026, 13 enacted that year, totaling 18 enacted laws across 2025–2026—and documents a shift from early disclosure/crisis-referral rules toward regulating conversational UX (crisis protocols, minor safeguards, emotional-dependence and engagement-optimization limits, professional-licensure boundaries, data protections). Notes chatbot rules may preview agentic-AI governance. Complements same-week Oregon/Maryland frontier procurement moves by showing parallel state activity on conversational-AI product rules.
Why it matters, evidence and limitations
- Why it matters
- Cross-state legislative intelligence for SLED sellers and state counsel: clarifies that companion/youth chatbot statutes may or may not sweep ordinary government customer-service bots depending on definitions/carveouts—transferable caution when deploying resident assistants.
- Evidence and measured results
- FPF desk analysis of introduced/enacted bills; not an empirical study of chatbot harm rates. Companion-chatbot focus is primary; government FAQ bots may fall into carveouts depending on state definitions.
- Limitations and uncertainty
- publicationDate set to 2026-09-01 as month-level; exact day not stated on cover. Counts are FPF-tracked through September 2026 and may move. Report is not an audit of state agency systems.
How to read this edition
Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.
- Government evaluation
- A public body’s measured evaluation or documented pilot.
- Independent research
- Research conducted outside the implementing organization.