From the State Government edition of October 4, 2026
Vermont Rep. Sibilia testimony (Sep 30): legislators learned of Anthropic cyber-AI deal by accident; $100K credits, $1,000 liability cap, AI ethics code not updated
Vermont General Assembly, Joint Information Technology Oversight Committee (testimony of Rep. Laura Sibilia) · State legislative IT oversight / agency use of frontier AI for defensive cybersecurity · Vermont statewide
- Publisher
- Vermont General Assembly (JITOC)
- Original publication
- 2026-09-30
- Source retrieved
- 2026-10-05
- Event date
- 2026-09-30
What happened
Rep. Laura Sibilia's written Sep 30 JITOC testimony says she first learned at the Aug 4 Vermont Digital Government Summit that the Agency of Digital Services (ADS) was in Anthropic's public-sector (SLTT) version of Project Glasswing, and that a state official called it fortunate Vermont did not have to go through the Legislature. She reports being twice told participation was not public, being shown an unsigned Anthropic Evaluation Agreement on Aug 28, and never receiving one requested PowerPoint. Legislative Counsel found no general requirement for legislative approval but pointed to existing AI, procurement and reporting laws and a statutorily required AI Code of Ethics that had not been updated annually. She questions using $100,000 in credits to beta test on state infrastructure under an agreement she says carried a $1,000 liability limit, and says a follow-up memo asks about confidential-data protection, contracting review and whether a NIST impact and risk assessment was done.
Why it matters
Directly on point for state CIOs and legislatures: a state agency adopted a vendor-subsidized agentic cyber-AI program under existing executive authority, and the oversight committee is now testing whether AI reporting, procurement and ethics-code statutes were followed. Why now: this primary testimony was TLS-inaccessible on our earlier run (legislature.vt.gov host) and opened successfully tonight via legislature.vermont.gov; the archive holds only the Sep 29 Vermont Public news item, not the legislative record.
Evidence and measured results
Opened the JITOC PDF over HTTPS (curl, UA SLED-research-bot/1.0) and extracted text with pdftotext; confirmed dates (Aug 4 summit, Aug 28 meeting, Sep 30 hearing), the $100,000 credits and $1,000 liability-limit statements, the unsigned evaluation agreement, the AI Code of Ethics not updated annually, and the three memo focus areas. Cross-checked the Sep 30 agenda (Anthropic SLTT Program briefing by Secretary Reilly-Hughes) and Sen. Plunkett's Sep 25 committee email describing the memo.
Limitations and uncertainty
Advocacy testimony by one legislator; ADS's own testimony and documents were not inspected. The $1,000 liability figure refers to the agreement she was shown, not necessarily final signed terms. No measured security outcomes are reported. Vermont-specific statutes limit transferability.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-04; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- State agencies are joining vendor-subsidized AI trials without a disclosure path to oversight bodies, creating legitimacy and contract-risk problems.
- Stakeholders
- Agency of Digital Services secretary
- State CISO
- Legislative IT oversight committee
- Legislative counsel and Joint Fiscal Office
- Procurement office
- Discovery questions
- How are free or credit-funded AI trials reviewed and recorded?
- Who tells the legislature about agentic tools on state systems?
- Is your statutory AI code of ethics or inventory current?
- What liability terms did you accept for beta AI services?
- Value hypothesis
- A short AI-agreement governance review can reduce oversight friction and surface liability and data-handling gaps before a trial expands.
- Potential engagement
- Two-to-three week review of one AI trial agreement, its approvals and impact assessment, ending in a disclosure package for oversight staff.
- Unsupported claims
- That the Vermont program caused any security incident
- That vendor tools reduced vulnerabilities by a measured amount
- That legislative approval was legally required
- That a governance review guarantees legislative acceptance
Pre-sales engineering
Role takeaway
- Fit
- Control design and documentation for agentic cyber tools operating on state code, not deployment of the tool itself.
- Architecture and integration
- Map in-scope repositories and systems, permission boundaries, data egress and vendor retention, and log every tool action into the agency SIEM.
- Prerequisites
- Current application and AI-use inventory
- Signed agreement and data-handling terms
- NIST AI RMF or equivalent impact-assessment template
- Named human approver for code changes
- Constraints
- Testimony lacks technical environment details
- Vermont-specific statutory duties
- Agreement terms seen were unsigned
- Vendor model and terms may change during the trial
- Security
- Treat the vendor tool as a privileged identity: least-privilege credentials, isolated environment, no production secrets in prompts, and human approval before any patch merges.
- Proof of value
- Written access map approved by CISO and shared with oversight staff
- Impact assessment matched to the configuration actually running
- Audit log sample showing every tool action is attributable
Delivery
Role takeaway
- Work
- Assemble the AI-trial records, complete an impact and risk assessment, update the AI ethics code, and brief the oversight committee.
- Dependencies
- Agency legal review of the agreement
- Legislative counsel input
- Vendor responses on data retention and model versions
- Ownership
- Agency of Digital Services secretary and CISO, with legislative counsel as reviewer.
- Skills and adoption
- Agency leaders need training on when AI trials trigger reporting, procurement and ethics-code duties.
- Governance checkpoints
- Pre-trial disclosure to oversight committee
- Impact assessment sign-off
- Annual AI ethics code update
- Liability and indemnity review
- Acceptance criteria
- Oversight committee receives full agreement and approvals within 30 days of request
- Impact assessment on file before scope expansion
- AI ethics code updated within the statutory cycle
- Risks
- Trial continues while oversight questions remain unanswered
- Vendor default liability terms left in place
- Stale inventory hides other AI agreements
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 program covers tools that scan and can patch state code, so the architecture question is scope: which repositories and systems the vendor tool can read, write and run against, and what leaves the state boundary. Testimony does not describe the technical environment; that remains the agency's to document.
Governance
Who approves, reviews and stays accountable for outcomes?
Shows the gap between 'no legislative approval required' and compliance with existing AI inventory, reporting, procurement and ethics-code duties. Oversight bodies are asking for the full agreement, approval and procurement records, and evidence of a NIST-style impact assessment before such pilots proceed.
Security and privacy
What data, permissions and controls need testing?
Questions center on confidential information, credentials and security information exposed to a beta frontier cyber tool, and on accountability if something goes wrong under a low liability cap. No incident is alleged.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Limited direct workforce impact; the testimony stresses that legislators and staff lacked visibility, implying training and disclosure duties for agency leaders on AI agreements.
Procurement
What should contracts, pricing and exit terms secure?
Vendor credit and evaluation agreements can bypass ordinary procurement scrutiny. States should route free or credit-funded AI trials through the same review, liability negotiation and records retention as paid contracts.
Operating model
Which teams own the service once it runs?
Assign a named owner for AI-agreement disclosure to oversight committees, and keep the statutory AI ethics code and inventory current so trials can be measured against them.
Publication history
- 2026-10-04State Government · Issue 285 resources
Stable resource ID: vermont-sibilia-glasswing-testimony-20260930