State & Local GovernmentIssue 332 records
State Government
Tonight’s edition is Vermont’s formal oversight packet on its Anthropic SLTT trial. ADS’s Oct 7 answers say the six-month code-review evaluation skipped Bulletin 3.5 procurement treatment at signing but will use Bulletin 3.5 waivers for future trials, is not using Claude Security or Mythos, runs Claude Code in VS Code on Azure-hosted models with human release review, and has used about 10% of a $100,000 credit. The signed evaluation agreement posted the same day still names Claude Security among the Evaluation Services—showing why authorized-product lists need to sit beside vendor forms.
- Evidence records
- 2
- Cross-source patterns
- 1
- Evidence classes
- 2 government evaluation
- Outcomes
- 1 mixed1 emerging
- Source freshness
- 2 new this fortnight
- Research completed
- 2026-10-09
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Evaluation agreements can name AI security products the agency later says it is not using
Vermont’s posted Anthropic form lists Claude Security among Evaluation Services, while ADS’s same-day JITOC answers state the state is not using Claude Security and lacks Mythos access. Oversight packets need an explicit authorized-product delta, not only the vendor form.
Operating questionWhen an AI evaluation contract lists multiple products, how does the agency record which ones are actually enabled?
Supporting evidenceVermont Agency of Digital Services (Secretary/CIO Denise Reilly-Hughes; Chief Data and AI Officer Josiah Raiche)State of Vermont and Anthropic, PBC (agreement posted by ADS to JITOC)
Evidence ledger
Full record · every source keeps its link and limitations
Vermont ADS tells JITOC it will apply Bulletin 3.5 waivers to future AI trials, is not using Claude Security or Mythos, and has used about 10% of a $100,000 Anthropic credit
On Oct 7, 2026, ADS posted formal answers to the Joint Information Technology Oversight Committee on its June 2026 Anthropic SLTT evaluation. Secretary/CIO Denise Reilly-Hughes says ADS entered a six-month evaluation to examine state system code for security and operational weaknesses; the evaluation is limited to system code and no personal information or Vermonters’ data has been reviewed. ADS did not bid, request a Bulletin 3.5 waiver, or treat the agreement as a procurement at signing, but after consulting the Secretary of Administration it will use the Bulletin 3.5 waiver process for trials and evaluations going forward. CDAO Josiah Raiche writes Vermont is not using Claude Security and does not use or have access to Claude Mythos; developers use the Claude Code plugin in VS Code with Sonnet, Opus, or Fable by application size, after a two-day in-person training. Sonnet/Opus prompts are not retained; Fable retains prompts and attached code up to 30 days for misuse review. All generated code goes through standard release review before production. ADS is using the Azure-hosted LLM, has used about 10% of the $100,000 credit allocation, and will add the use case to the state’s AI Inventory at the evaluation’s conclusion. Authorization was signed by the then-serving CISO and reviewed by ADS General Counsel and the CDAO.
Why it matters, evidence and limitations
- Why it matters
- Primary-source answers to legislative oversight on a free frontier-AI cybersecurity trial: procurement classification, model choice, data retention, human release review, and a prospective process change (Bulletin 3.5 waivers for future trials). Transferable as a template for how other states document SLTT-style evaluations under legislative scrutiny, with Vermont-specific Bulletin 3.5 and inventory rules as jurisdictional limits.
- Evidence and measured results
- Opened the Oct 7 ADS Response PDF from the JITOC documents page (Denise Reilly-Hughes cover letter, Raiche memorandum, Attachment A Q&A dated Oct 6) via curl and pdftotext. Confirmed listing on https://legislature.vermont.gov/committee/document/2026/345/Date/10-7-2026. Cross-checked Compass Vermont reporting (already archived) and the signed evaluation agreement posted the same day.
- Limitations and uncertainty
- Agency self-report under oversight pressure, not an independent audit. No count or severity of vulnerabilities found, no measured remediation outcomes, and no confirmation whether Fable was used on state code. Attachment A is dated Oct 6 while the cover letter is Oct 7. The agreement title/services include Claude Security even though ADS says it is not used—see companion contract resource.
Vermont posts the signed Anthropic Evaluation Agreement: $100,000 credit for Claude Security and Claude Code, defensive cyber use only, six-month term
Alongside its Oct 7 answers, ADS posted the signed Anthropic Evaluation Agreement titled “Claude Security (Government).” Customer is the State of Vermont. Evaluation Services are “Claude for Enterprise including Claude Security and Claude Code with cyber defense runbooks,” with a Credit Amount of $100,000.00 USD for a six-month Evaluation Period from the Effective Date. Approved use is solely for defensive cybersecurity purposes in customer-owned or authorized systems (scanning, triaging, validating, remediating vulnerabilities); general software development is barred unless Anthropic permits in writing. Services are provided “as is”; Claude Security is labeled a Beta Service; liability is capped at $1,000 or fees paid. Government terms allow public-disclosure-law sharing of the agreement with notice to Anthropic. This contract text sits beside ADS’s same-day statement that Vermont is not using Claude Security.
Why it matters, evidence and limitations
- Why it matters
- Shows the actual instrument behind a free state SLTT AI security trial: credit amount, scoped defensive uses, beta disclaimer, and liability cap. Useful for peer states comparing vendor evaluation templates. Jurisdictional limit: Vermont’s Bulletin 3.5 treatment is separate from the vendor form.
- Evidence and measured results
- Opened the PDF posted on the JITOC Oct 7 documents page via curl and pdftotext. Compared service description to ADS’s written statement that Claude Security is not used.
- Limitations and uncertainty
- Effective Date and wet-ink signature dates are not clearly extractable from the text layer (form fields). Title and Evaluation Services include Claude Security even though ADS says it is not using that product—scope on paper may exceed stated practice. Liability and as-is terms are vendor-favorable; no measured security outcomes appear in the contract.
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.