WeVote
Log In

Bill

Bill

H 341

An act relating to creating oversight and safety standards for developers and deployers of inherently dangerous artificial intelligence systems

2025-2026 Regular Session Introduced by Angela Arsenault and 8 co-sponsors

Vermont bill would create a regulatory framework to identify, monitor, and mitigate risks from inherently dangerous AI, with safety standards, oversight, and penalties.

Read first time and referred to the Committee on Commerce and Economic Development
0
WeVote Research Nonpartisan
Bill Summary · H 341

Overview

H 341 (2025-2026) from Vermont proposes creating oversight and safety standards for developers and deployers of inherently dangerous artificial intelligence systems. The bill aims to establish a regulatory framework to identify, monitor, and mitigate risks associated with AI systems deemed inherently dangerous, with specific requirements for transparency, safety practices, accountability, and enforcement.

Main purpose and intent

  • Set up a formal oversight structure to manage risks posed by inherently dangerous AI systems.
  • Establish safety standards to reduce potential harm to the public, consumers, workers, and infrastructure.
  • Create requirements for developers and deployers to ensure responsible design, testing, deployment, and ongoing monitoring of such AI systems.
  • Provide a mechanism for reporting, investigation, and penalties for noncompliance.

Key provisions and changes

  • Definitions: The bill defines what constitutes an “inherently dangerous artificial intelligence system,” clarifying the scope (e.g., high-risk autonomous decision-making, critical infrastructure involvement, safety-critical applications, or systems with significant potential for harm).
  • Oversight Authority: Establishes a government oversight entity or designates a relevant state agency (likely Commerce and Economic Development or a parallel body) to regulate these AI systems.
  • Safety Standards:
    • Mandatory risk assessment and impact assessments prior to deployment.
    • Required due diligence on safety by design, robustness, reliability, and resilience.
    • Mandatory post-deployment monitoring and incident reporting.
    • Requirements for human oversight, fail-safes, and rollback capabilities where appropriate.
  • Registration and Licensing:
    • Developers and deployers of inherently dangerous AI would need to register with the oversight authority.
    • Possible licensing or certification processes for certain high-risk AI deployments.
  • Transparency and Disclosure:
    • Obligations to disclose high-level information about the AI system’s purpose, capabilities, limitations, and data governance practices.
    • Documentation of safety assurances and testing results.
  • Data Governance:
    • Standards for data quality, provenance, privacy protections, and mitigation of bias.
  • Accountability and Governance:
    • Clear liability framework for harms caused by these systems.
    • Mechanisms for audits, whistleblower protections, and stakeholder input.
  • Penalties and Enforcement:
    • Administrative penalties, fines, license suspensions or revocations for noncompliance.
    • Possible civil or administrative remedies for affected parties.
  • Public and Consumer Protections:
    • Whistleblower protections and avenues for public complaints.
    • Requirements for accessibility of safety information to the public in certain cases.
  • Reporting and Review:
    • Regular reporting deadlines to the legislature or oversight body.
    • Five-year or periodic sunset/renewal provisions to reassess regulatory scope (if included).

Who would be affected

  • AI Developers and Vendors: Entities that design, build, test, and commercialize inherently dangerous AI systems.
  • Deployers and Operators: Organizations that implement such systems in real-world settings (e.g., critical infrastructure, public services, or safety-critical industries).
  • Regulated Industries: Sectors involving safety-critical decisions or high-risk outcomes (e.g., transportation, energy, healthcare, public safety) may be subject to additional compliance requirements.
  • General Public and Consumers: Beneficiaries of increased transparency, safety assurances, and recourse mechanisms against harmful AI deployments.
  • State Agencies: Affected agencies would gain new regulatory authority, registration duties, and enforcement responsibilities.

Procedural and timeline aspects

  • Introduction and Referral: The bill was read in the House and referred to the Committee on Commerce and Economic Development on February 25, 2025.
  • Legislative Process: As a first-step measure, the committee would review, possibly amend, and move the bill through hearings, votes, and potential revisions before floor consideration.
  • Regulatory Timeline: If enacted, implementation timelines would typically include phased compliance (e.g., a grace period for registration, deadlines for safety assessments, and a start date for oversight enforcement).
  • Ongoing Oversight: Provisions likely provide for ongoing oversight, annual or biennial reporting, and potential sunset or renewal provisions to reassess the program’s scope.

Potential impacts and considerations

  • Strengthened safety and accountability for high-risk AI deployments.
  • Increased compliance costs and administrative burden for developers and deployers.
  • Greater transparency for public-facing AI systems and potential data governance enhancements.
  • More robust mechanisms to mitigate harm, bias, and unintended consequences of dangerous AI systems.
  • Possible need for interagency coordination and alignment with federal or international AI safety standards.

Note: This summary is based on the bill’s title, description, and the listed sponsor information. For precise statutory language, definitions, scope, and enforcement details, consult the official bill text and fiscal notes once released by the Vermont General Assembly.

Compiled from official sources — confirm details with the bill’s official record.

Sign in to ask a question.