AI Risk Plans for Critical Infrastructure

Summary#

HB 1521, the Virginia Digital Innovation and Infrastructure Act, would add a new chapter to Title 59.1 of the Code of Virginia. Its stated goal is to protect rights in technology and digital property, set risk rules for some AI use at critical infrastructure facilities, and create uniform statewide rules. The supplied status is “introduced” and “continued from last session,” so the bill is not law.

  • It would require government restrictions on people’s lawful use of technology resources to meet a strict legal test: they must serve a compelling government interest and be narrowly tailored as the least restrictive way to achieve it.
  • It would treat digital assets as personal property and list owners’ rights to control access, transfer ownership, delete or modify assets where other law allows, use technology resources lawfully, and be secure against unreasonable searches and seizures.
  • A government entity could compel disclosure of a key, password, or other security credential only under a warrant based on probable cause. Law enforcement could still seize property under a warrant or a recognized warrant exception; permanent forfeiture would generally require a related criminal conviction, with federal-law exceptions preserved.
  • A business that deploys a defined “critical artificial intelligence system” to control a defined critical infrastructure facility would have to keep and follow a risk-management policy, review it at least yearly, and file an annual attestation. A substantially equivalent federal risk plan would count.
  • The bill would limit local rules on technology, digital assets, and AI, while preserving local authority over matters such as zoning, physical impacts, general building and fire codes, and a local government’s own use of AI.

What it means for you#

  • People who own digital assets or technology resources: The bill would set out property and privacy-related rights, but it would not make those rights absolute. Other laws still apply, and the bill preserves law enforcement powers and federal law.
  • Businesses deploying certain AI systems at critical infrastructure facilities: They would have to develop, maintain, follow, and annually review a risk-management policy. They would also have to file a yearly attestation with the Secretary of Commerce and Trade. The bill defines covered facilities and systems; among other limits, a covered AI system must control facility operations without contemporaneous human oversight.
  • Those businesses and people bringing claims: A deployer meeting the bill’s conditions could use a state-law safe harbor against some damages claims. It would not cover personal injury or wrongful death, intentional or knowing violations of law, federal claims, or Attorney General actions. The bill does not create a private lawsuit for failing to follow the risk-policy requirement.
  • Local governments: They could not adopt or enforce certain local restrictions or extra AI-deployer requirements. They would retain the listed powers over physical impacts and their own AI use.
  • Consumers: The bill says it would not reduce rights under the Virginia Consumer Data Protection Act or the Uniform Fiduciary Access to Digital Assets Act.

Money#

No cost information is in the available material.

  • The bill would require covered deployers to prepare and maintain policies and file attestations. It gives no estimate of those compliance costs.
  • The Secretary of Commerce and Trade would have to maintain a public registry of attestations, with confidential information withheld. The bill states no staffing, funding, or cost estimate for that work.
  • The Attorney General could seek a civil penalty of up to $10,000 per violation for specified attestation failures or materially false attestations.

What is unclear#

  • The supplied bill text sets out proposed effective dates of July 1, 2026, for most provisions and July 1, 2027, for attestations. Since the supplied status is “introduced,” it is unclear when the rules would start if the bill were enacted later.
  • The bill does not set out a process for checking whether risk policies are adequate, or for verifying attestations beyond allowing the Attorney General to act in specified cases.
  • It does not explain how the credential-warrant rule interacts with the separate statement preserving subpoenas and other lawful process.
  • The bill uses terms such as “lawful purposes” and “materially false” without defining them here. No other current-law text or fiscal material was supplied, so overlap and costs beyond what this bill says cannot be verified.

Case for#

  • The bill appears intended to give people clearer property and privacy protections for digital assets and technology resources.
  • A statewide rule could prevent local requirements from differing across counties, cities, and towns.
  • Requiring risk policies for AI that controls critical infrastructure could prompt deployers to identify and address safety risks.
  • Allowing an equivalent federal risk plan to count could avoid requiring a second policy for substantially the same risks.

Case against#

  • The bill’s broad definition of technology resources and strict test for government restrictions could make some government rules harder to adopt or enforce.
  • Local governments would lose power to set certain AI and technology rules, even as the bill leaves some local authority intact.
  • The annual attestation is a deployer’s certification; the bill does not require an independent review of each policy or system.
  • The safe harbor could limit some state-law damages claims, while the bill creates no private action to enforce the risk-policy duty.