State & Local GovernmentIssue 603 records
State Government
Connecticut drew hard lines on candidate AI in hiring, and New York’s minor chatbot law is being read as a feature-level model. DAS General Letter 195 (Nov 4) lets executive-branch candidates use AI for research and polishing their own materials but bars AI-generated application answers, independent-work completion and live interview coaching, with a mandatory pause-and-document path for managers. Route Fifty’s Nov 6 analysis frames New York’s ban on unsafe companion features for under-18s—effective Jan 1, 2027—as feature-level regulation rather than a blanket youth AI ban, citing Common Sense Media testing gaps on ChatGPT for Teens and Florida’s proposed Safe Screens Act. Separately, Georgia’s (country) Parliament staff filed to buy ChatGPT Business for seven seats plus Claude AI Pro, awaiting procurement approval by Nov 10. House Oversight’s Delaware AIC document request and North Carolina’s GenAI RFP questions remain due around Nov 9.
- Evidence records
- 3
- Cross-source patterns
- 1
- Evidence classes
- 2 independent reporting1 standards or public-body guidance
- Outcomes
- 3 emerging
- Source freshness
- 3 new this fortnight
- Research completed
- 2026-11-07
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
States are writing feature- and role-specific AI rules instead of blanket bans
Connecticut’s DAS General Letter 195 lists permitted AI research/editing uses for job candidates and separately bans AI-generated application answers and live interview coaching, while Route Fifty’s Nov 6 analysis of New York’s minor chatbot law highlights expert support for feature-level companion restrictions over blanket bans on youth AI access. Both treat precise behavioral boundaries—not total prohibition—as the current governance move.
Operating questionWhere AI touches hiring or minors, does the state enumerate permitted versus prohibited behaviors and features, or only issue a general ban or permission?
Supporting evidenceConnecticut Department of Administrative Services (DAS), Statewide Human Resources Management DivisionRoute Fifty
Evidence ledger
Full record · every source keeps its link and limitations
Connecticut DAS General Letter 195 sets permitted and prohibited AI uses for candidates in executive-branch vacancy filling
Dated 4 November 2026, Connecticut DAS General Letter No. 195 tells executive-branch candidates under DAS Statewide HR that they remain solely responsible for accurate, truthful application content whether or not they use AI. Permitted uses are research on positions/agencies, interview practice, and grammar/organization/clarity edits of candidate-authored materials. Prohibited unless expressly authorized: generating answers to application or screening questions; completing exams, assessments, work samples or other independent-work activities; real-time interview assistance; and recording/transcribing interviews or introducing AI meeting assistants without advance authorization. Hiring managers who suspect unauthorized AI use must pause, describe the observation, hear the candidate, then decide whether to continue, require in-person activity, or end consideration, and document the decision to Agency HR. Talent Solutions must notice the policy on JobAps postings; hiring managers must reiterate it at invitation and before activities. Accessibility/ADA tools remain allowed. Why now: the letter is the operative statewide hiring rule as of 4 Nov and was absent from the 5 Nov state-government edition.
Why it matters, evidence and limitations
- Why it matters
- Direct U.S. state workforce control for GenAI in hiring. Transferable to other states writing candidate-side AI rules that complement employer-side bans (e.g., Kentucky’s earlier bar on AI in application review). Keep Connecticut executive-branch scope; does not change General Letter 194 process steps.
- Evidence and measured results
- Opened DAS page states Date: November 04, 2026; Purpose/Scope/Policy; named AI tools (ChatGPT, Claude, Copilot, Gemini, Teal, Jobscan, Final Round AI, LockedIn AI); Permitted Uses and Prohibited Uses lists; Response to Suspected Violations steps; Talent Solutions/JobAps and hiring-manager notice duties; ADA accommodation carve-out.
- Limitations and uncertainty
- Letter governs candidates, not agency use of AI to screen applicants. No measured incidence of AI-assisted applications or enforcement outcomes yet. Effective operational date is the letter date; compliance metrics are not published.
Georgia (country) Parliament staff seeks ChatGPT Business for seven seats and Claude AI Pro pending procurement approval
On 6 November 2026 SOVA reported that the Staff of the Parliament of Georgia plans an annual ChatGPT Business subscription for seven employees at up to 10,500 GEL, plus 826.84 GEL for Claude AI Pro, with details published on the state procurement website. Intended uses include routine data handling and document processing; the Parliamentary Budget Office expects ChatGPT for statistics, analytical reports and economic forecasts. Justification cites workflow optimization and avoiding intermediary markups via direct subscription. The request originated with the Deputy Head of Parliamentary Staff, Public Relations, and the Budget Office; an application to the State Procurement Agency was filed 4 November with a decision expected by 10 November. No measured productivity results are claimed.
Why it matters, evidence and limitations
- Why it matters
- Transferable international evidence for small-seat legislative/staff GenAI procurement under a formal purchase approval path. Jurisdictional limits: national parliament of Georgia (country), not a U.S. state; currency and procurement-agency process differ; treat seat counts and GEL ceilings as local filing details, not U.S. benchmarks.
- Evidence and measured results
- Opened SOVA article states seven ChatGPT Business seats up to 10,500 GEL, Claude AI Pro 826.84 GEL, Budget Office use cases, direct-subscription cost rationale, November 4 application and November 10 expected decision.
- Limitations and uncertainty
- Procurement still under review (decision expected Nov 10). Figures and use cases come from reporting on procurement filings, not an independent evaluation. No outcomes measured. bm.ge parallel report was HTTP 403 on curl; SOVA page opened successfully.
Route Fifty: experts say New York’s feature-level AI chatbot limits for minors may outperform blanket bans; Florida proposes similar Safe Screens rules
On 6 November 2026 Route Fifty reported that New York’s newly signed law barring unsafe AI companion features for people under 18—personified emotions, claimed personal relationships, self-harm encouragement, secrecy, sexual content and sycophancy—takes effect 1 January 2027 and is being framed by Brown University’s Shufang Sun and Common Sense Media’s Amina Fazlullah as feature-level regulation rather than a blanket ban on minors’ AI access. The article cites Common Sense Media Youth Safety Institute testing that found ChatGPT for Teens missed more than one in four warranted crisis referrals and failed expected parental alerts. It also notes Florida AG James Uthmeier’s late-October Safe Screens Act proposal with age determination and similar harmful-feature bans, plus California’s September companion-chatbot child-safety law and Tennessee’s ban on representing AI as qualified mental-health professionals.
Why it matters, evidence and limitations
- Why it matters
- State oversight of commercial AI products affecting residents, complementary to agency-internal GenAI governance. Transferable drafting pattern: prohibit specific interaction features and require age assurance rather than outlawing all minor access. Keep distinction from archived 28 Oct Hochul bill-signing primary; this is 6 Nov expert/comparative analysis. Not a K-12 school AI policy.
- Evidence and measured results
- Opened Route Fifty page dated November 6, 2026 03:43 PM ET; quotes Sun on feature-level vs blanket ban; Fazlullah on worst-case features and durable multi-platform scope; ChatGPT for Teens crisis-referral/parental-alert findings; effective date Jan. 1, 2027; Florida Safe Screens mention; California and Tennessee comparators.
- Limitations and uncertainty
- Underlying NY statute was signed 28 Oct and is already archived as new-york-s9051b-...; this resource adds 6 Nov expert framing and comparative state proposals, not new statutory text. ChatGPT for Teens test results are attributed to Common Sense Media, not independently re-run here. Florida measure is a proposal, not enacted.
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.
- Independent reporting
- Independent reporting with attributable sources but without a formal evaluation design.
- Standards or public-body guidance
- Normative or advisory guidance from a standards body or public institution.