AI Liability Framework for Developers

Summary#

This bill creates a federal product-liability law specifically for advanced artificial intelligence (AI) systems. It defines when AI developers and some users (called deployers) can be held legally responsible for harms from an AI system. The bill also bars certain contract terms that would unfairly limit liability and requires foreign AI developers to name a U.S. agent.

• Main change: A new, nationwide legal standard for civil liability when an AI system (called a “covered product”) causes harm.
• Who may be liable: Developers are liable for defective design, failure to warn, breach of express warranty, or if the product is unreasonably dangerous. Deployers can be treated like developers if they substantially modify a system or intentionally misuse it.
• Contract limits banned: Developers and deployers cannot include contract or terms-of-service language that waives or unreasonably limits liability under this law.
• Enforcement and remedies: The U.S. Attorney General, state attorneys general, individuals, or classes can sue in federal court for injunctions, damages, restitution, fees, and (for the U.S. AG) civil penalties.
• Foreign providers: Foreign developers must designate a U.S. resident agent before making their AI available in the U.S.; the Attorney General keeps a public registry.
• Timing and suits: Lawsuits must generally be filed within four years of discovering the harm and its cause. The law applies to any suit filed on or after enactment, even if the harm occurred earlier.

What it means for you#

• Developers (companies and researchers who design or produce AI):

  • May face liability if a court finds they failed to use reasonable care in design, warnings, or if the product was unreasonably dangerous.
  • Strict liability applies for harms caused by a defective condition that made the product unreasonably dangerous, even if the developer exercised all possible care.
  • Cannot use contracts with deployers to waive or unreasonably limit liability.
  • If the developer is foreign, they must name a U.S. resident agent and keep that information current.

• Deployers (businesses or people who use or offer AI services to others):

  • Generally are not treated as developers unless they substantially modify the AI or intentionally misuse it.
  • If the developer cannot be sued (out of jurisdiction or insolvent), a deployer can be held responsible to the same extent as the developer.
  • Deployers who are held liable may seek indemnification (payment of costs) from developers, but recovery may be limited if the deployer was partly at fault.

• Consumers and people harmed by AI systems:

  • Can sue developers or deployers in federal court for injury, financial loss, reputational harm, or severe emotional distress caused by an AI system.
  • Must file a suit generally within four years after discovering the harm and its cause.

• State and federal governments:

  • State attorneys general and the U.S. Attorney General can enforce the law and seek injunctive relief, damages, and penalties.
  • The law supersedes state law only where state law conflicts; states may keep stronger protections.

• Foreign AI companies:

  • Must designate a U.S. resident agent before offering AI products in the U.S. or risk being blocked from deployment.

Expenses#

No publicly available information.

• The bill does not include a fiscal note in the provided text.
• This could increase litigation and enforcement costs for the federal government and for state attorneys general, because it creates a federal private cause of action and an enforcement role for the U.S. Attorney General.
• Developers and deployers may face higher compliance, legal, and insurance costs to manage new liability risks and to meet warning, documentation, and design standards.
• Maintaining the public registry and processing agent designations will likely require staff time at the Department of Justice.

Proponents' View#

• The bill appears intended to create clearer legal rules for harms from AI systems. Supporters may argue that clearer rules will:
  • Encourage safer AI design by making developers accountable for foreseeable risks.
  • Provide more predictable legal outcomes for businesses and consumers, reducing uncertainty that could slow innovation or interstate commerce.
  • Protect individuals and small businesses from harms caused by AI.
  • Promote transparency and accountability for foreign AI providers by requiring a U.S. agent.

Opponents' View#

• One concern is that the bill could increase litigation costs and legal uncertainty for AI developers, especially over broad or unclear terms such as “design,” “substantial modification,” and what counts as “reasonable care.”
• The rule that lawsuits filed after enactment can cover harms that happened earlier could expose developers to unexpected claims for past conduct. The text makes suits prospective by filing date, which may surprise some companies.
• Requiring foreign developers to name a U.S. resident agent and blocking deployment for noncompliance could limit access to foreign AI services or shift costs to small providers.
• The law leaves room for state laws that are stronger; this mixed federal-state approach could produce patchwork rules and complexity for companies operating in many states.
• It is unclear how courts will apply strict liability in practice to software-based systems and whether existing insurance markets will adapt easily to cover these risks.

If you want, I can summarize specific sections line-by-line or create a short Q&A about how a developer or small business should respond if this law becomes enacted.