Protects state and local governments’ authority to adopt stronger artificial intelligence safeguards.

• Establishes federal AI protections as a minimum standard rather than a national ceiling.

• Prohibits federal preemption of stronger state and local civil-rights, safety, privacy, or accountability laws.

• Bars federal retaliation against communities that regulate, restrict, or prohibit dangerous AI systems.

Requires AI companies to prevent foreseeable harm before releasing or deploying systems.

- Requires documented risk assessments and reasonable safeguards before consequential systems are released.

- Requires companies to monitor deployed systems and correct newly discovered dangers.

- Treats failure to address known or reasonably foreseeable risks as a breach of legal duty.

Creates a clear legal right to recover damages caused by artificial intelligence.

• Allows people to sue developers, deployers, and operators responsible for harmful AI systems.

• Covers physical injury, financial loss, property damage, civil-rights violations, and other material harm.

• Authorizes compensation, corrective orders, injunctions, and additional damages for reckless or intentional conduct.

Prevents companies from receiving special immunity for injuries caused by artificial intelligence.

• Prohibits federal immunity written specifically to protect AI developers, vendors, or deployers.

• Prevents companies from escaping responsibility by claiming an AI system acted independently.

• Invalidates contract terms and disclaimers that attempt to eliminate legally required accountability.

Requires companies to preserve records needed to investigate consequential artificial intelligence decisions.

• Preserves relevant inputs, outputs, system versions, testing results, instructions, and decision records.

• Establishes retention periods based on the seriousness and duration of possible harm.

• Imposes meaningful penalties when companies conceal, alter, destroy, or fail to preserve evidence.

Guarantees human review before automated systems deny essential public assistance.

• Covers Social Security, disability assistance, nutrition benefits, health coverage, housing, and unemployment programs.

• Requires written reasons, supporting evidence, and an accessible opportunity to challenge adverse decisions.

• Continues existing assistance during timely appeals unless fraud or immediate public danger is established.

Prohibits artificial intelligence from independently determining liberty, sentencing, or pretrial release.

• Prohibits automated systems from independently determining bail, sentencing, incarceration, probation, or parole.

• Gives defendants access to relevant data, validation results, error rates, and bias assessments.

• Requires judges to make independent findings and provide defendants an opportunity to challenge algorithmic evidence.

Keeps qualified medical professionals responsible for decisions affecting patient care.

• Prohibits insurers and providers from denying or delaying care solely through automated decisions.

• Protects medical professionals who override AI recommendations based on their independent clinical judgment.

• Gives patients notice, understandable explanations, and expedited human appeals for adverse decisions.