WeVote
Log In

Bill

Bill

HB 934

Artificial Intelligence Regulatory Reform Act.

2025-2026 Session Introduced by Eric Ager and 12 co-sponsors

The bill creates a new criminal and civil remedy for creating or distributing deepfakes without consent, while immunizing AI developers when used by licensed professionals.

Reptd Fav Com Substitute
#computers;courts;crimes;emergingtechnologies;fines&penalties;fraud;informationtechnology;obscenity;photography;public;stem;liability;civilactions;artificialintelligence
0
WeVote Research Nonpartisan
Bill Summary · HB 934

HB 934 — “AI Regulatory Reform Act” (House Bill 934)

Status: Reported Favorably as Committee Substitute (Reptd Fav Com Substitute)
Introduced: Nov 12, 2024 (multiple drafting/edition dates shown)
Primary sponsor (NC version): Rep. Johnson
Primary subject areas: artificial intelligence, deepfakes, criminal law, civil liability, elections, professional services

Purpose / intent

The bill has two principal aims: (1) create a new criminal and civil cause of action targeting malicious creation and distribution of “deepfakes,” and (2) limit civil liability for developers of AI products when licensed or “learned” professionals use those AI tools in the course of providing professional services to clients. The stated policy objectives are to protect individuals and elections from harmful synthetic media while clarifying legal responsibility for AI-driven errors in professional settings.

Key provisions

1) New offense — “Unlawful distribution of a deepfake”

  • Definition: “Deepfake” = image, audio, or video digitally created or altered with intent to deceive that inauthentically depicts a real person speaking or acting in ways they did not (bill versions include a standard that a reasonable person would not know the depiction is false). Works of artistic or newsworthy value (commentary, criticism, satire, parody) are excluded in committee text.
  • Prohibited conduct (without affirmative consent of the depicted person), when committed for purposes such as harassment, extortion, threats, causing physical/emotional/reputational/economic harm, or injuring a political candidate or influencing an election:
    • Creating a deepfake with intent to distribute;
    • Distributing a deepfake;
    • Soliciting creation of a deepfake with intent to distribute.
  • Criminal penalty: Class 1 misdemeanor.
  • Remedies: Court may order destruction of the offending deepfake.
  • Civil remedies: Aggrieved persons may bring a civil suit and recover actual damages (but not less than liquidated damages computed at $1,000 per redistribution or $10,000, whichever is higher), punitive damages, and reasonable attorneys’ fees.
  • Exceptions in committee substitute: interactive computer services and providers/developers of technology used to create deepfakes are (in some drafts) carved out from liability for third-party user conduct.

Effective date (drafted): provisions apply to offenses committed on or after Dec 1, 2025 (if enacted as drafted).

2) Civil immunity for AI developers when used by “learned professionals”

  • Definitions provided: “AI product,” “developer,” “learned professional” (licensed professionals bound by standards and exercising independent professional judgment), “client,” and “error” (objectively incorrect output or failure to perform represented function).
  • Core rule: When a learned professional uses an AI product in delivering services to a client, the developer of the AI product is not liable for errors produced by the product; the learned professional is solely responsible for any client harm resulting from those AI-generated errors.
  • Additional drafting language (committee versions) clarifies interactive computer services are likewise not liable for a learned professional’s use of an AI product.
  • Other immunities: the Article does not abrogate other statutory or common-law immunities.

Effective date (drafted): applies to acts/omissions on or after Dec 1, 2025 (if enacted as drafted).

Who is affected

  • Individuals targeted by malicious synthetic media (potential plaintiffs and victims).
  • Creators, distributors, and requesters of deepfakes (potential criminal defendants and civil defendants).
  • AI developers, platform providers, and intermediaries (protected from certain liabilities in professional-use contexts under the bill’s immunity provisions; some drafts also exempt platforms for third‑party deepfake uploads).
  • Learned professionals (e.g., licensed lawyers, medical providers, accountants) who use AI tools — they would bear primary civil responsibility for errors arising from AI outputs used in client services.
  • Election candidates and campaigns (explicitly covered as protected categories in deepfake provisions).

Procedural / timing notes

  • Several draft editions and committee substitutes exist; the committee substitute includes expanded definitions and carve-outs (e.g., for interactive computer services and expressive uses).
  • Drafts indicate the criminal and civil provisions would take effect Dec 1, 2025 and apply to acts on or after that date.
  • The version labeled “Committee Substitute Favorable” indicates the bill has been favorably reported by committee with amendments (status: Reptd Fav Com Substitute).

Potential impacts and considerations

  • Strengthens remedies against malicious synthetic media by combining criminal penalties, statutory civil damages, and destruction orders.
  • Shifts legal risk for AI errors in professional contexts from developers to the licensed professionals who deploy AI, potentially reducing developers’ liability exposure but increasing professional duty to validate AI outputs.
  • May raise questions about access to recourse for clients harmed by opaque or complex AI systems (practical ability of professionals to detect errors, and availability of recovery when professionals lack means).
  • Interaction with platform liability regimes and free-speech/expression exceptions (news, satire) will be legally significant; committee language attempts to preserve certain expressive uses.
  • Implementation and litigation will clarify coverage boundaries (e.g., what qualifies as a “learned professional,” what constitutes affirmative consent, how “intent to deceive” will be proved).

If you want, I can:
- produce a one-page bill comparison showing differences among the draft editions (original, committee substitute, conference versions), or
- produce a short explainer focused on likely compliance steps for AI developers and learned professionals.

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

Sign in to ask a question.