H.R. 10841 — Securing Elections From AI Deception Act

This section establishes definitions for the Act, including the Federal Trade Commission as the Commission; a covered algorithm as any computational process using machine learning, artificial intelligence, or similar techniques that creates products or information, affects content display, or supports decisions; a deployer as any non-individual person using such an algorithm; a developer as any person designing, coding, or substantially modifying such an algorithm; and an election worker as an official, poll worker, or volunteer administering federal, state, or local elections.

This section prohibits developers or deployers from using a covered algorithm to intentionally deprive or defraud individuals of the right to vote in federal, state, or local elections. It bars providing deceptive information on voting times, places, eligibility, ballot counting, endorsements, or other procedures, and using deception, threats, intimidation, fraud, or coercion to interfere with registering, voting, supporting candidates, or serving as election workers. Violations carry fines under title 18, up to one year imprisonment, or both, effective 60 days after enactment.

This section requires deployers using covered algorithms to generate and intentionally publish images, videos, audio, or other media simulating or altering candidates for federal, state, or local office or depicting voting or voter registration information to include either the statement “AI Disclaimer: This output was generated by artificial intelligence” or an “AI” symbol. It directs developers to provide deployers a compliance mechanism and, to the extent technically feasible, make disclaimers permanent. The Federal Trade Commission must issue implementing regulations within 2 years, enforce violations as unfair or deceptive acts, conduct public education for 5 years, and report to Congress on effectiveness. The requirements take effect 60 days after the regulations are issued.

This section imposes a duty of care on developers and deployers of covered algorithms by prohibiting their offer, license, or use unless the algorithms are safe and effective. It defines safe algorithms as those for which reasonable harm-mitigation measures are taken, intended use complies with the Act, and a four-part balancing test shows no substantial individual harm, net benefits, reasonable avoidance, and no deception. It defines effective algorithms as those meeting ordinary skill performance standards, advertised performance and purpose, and using relevant data. It treats violations as unfair or deceptive acts enforceable by the Federal Trade Commission under the Federal Trade Commission Act.