Collusive Algorithms Act bans pricing AI

Summary#

SB-1507, the proposed Collusive Algorithms Act, would create a new Pennsylvania law; it does not amend an existing Act. Its stated goal is to prevent algorithmic pricing from producing artificially high prices and collusion-like results. The bill is introduced and is not law.

  • Businesses could not distribute or develop for distribution a “collusive algorithmic pricing structure”—one that uses or was trained on nonpublic competitor data to recommend or set prices or commercial terms in a relevant antitrust market.
  • A person could not use such a structure to recommend a price or commercial term. For this use alone, there is an exception if substantial evidence shows the person did not know the structure used or was trained on that data.
  • A person could not use “common artificial intelligence” (AI not used exclusively by one person) to set prices for goods or services offered or sold in Pennsylvania. This ban has no stated exception for lack of competitor data.
  • The Attorney General may investigate and prosecute violations, seek court orders to stop them, accept written voluntary-compliance agreements, and make regulations. A court may order restitution after issuing a permanent injunction.
  • Consumers who bought goods or services and suffered a measurable loss because of a violation could sue. A court may award damages, attorney fees and court costs, and other relief.

What it means for you#

  • Businesses and AI developers: The bill would ban certain development, distribution and use of pricing systems involving nonpublic competitor data. It would also bar using shared AI to set prices, even if the system does not use competitor data. “Nonpublic data” includes some public-facing data offered under terms of service that prohibit its use.
  • Consumers: You could bring a private case only if you bought goods or services and suffered a measurable loss because of a violation. A court could award actual damages or at least $1,000, whichever is greater, and could award up to three times actual damages, but not less than $1,000.
  • Attorney General and courts: The Attorney General may bring enforcement actions, but the bill does not require the office to pursue every suspected violation. A court may order restitution of money or property acquired through a violation after a permanent injunction.
  • People generally: The bill would not set prices or guarantee lower prices. It would not generally ban pricing algorithms or single-user AI that do not use or were not trained on nonpublic competitor data, except for the separate ban on shared AI setting prices.

Money#

No cost estimate is in the supplied material.

  • The bill authorizes Attorney General enforcement and allows regulations, but does not require a particular staffing level or provide an appropriation.
  • Criminal fines could be up to $1 million for each distribution, $100 million for each collusive pricing structure developed for distribution, and $100,000 for each pricing decision or recommendation using one. Using common AI to set a price could bring a fine of up to $5,000 per pricing decision or recommendation. The bill labels these offenses by felony or misdemeanor degree.
  • Courts could award consumer damages, attorney fees and costs, other relief, and restitution in the circumstances the bill describes. The bill gives no estimate of these amounts or says where fines would go.

What is unclear#

  • “Commercial term” is not defined. The bill refers to federal merger guidelines to identify a relevant antitrust market, but does not explain how that definition would be applied in these cases.
  • The bill does not set rules for investigations, evidence, hearings, appeals or how businesses can determine whether a system used nonpublic competitor data.
  • “Common artificial intelligence” could cover systems used by more than one person, but the bill does not clarify how shared use would be established.
  • The bill does not explain how its rules would interact with existing Pennsylvania law. It says federal antitrust remedies remain available.
  • The supplied material does not describe current Pennsylvania law on this subject, so a comparison with existing state rules cannot be verified.

Case for#

  • The bill appears intended to address the risk that businesses using the same pricing system could produce similar, elevated prices, including without knowingly coordinating.
  • Restricting systems that use nonpublic competitor data could reduce the risk that confidential business information feeds into competitors’ pricing.
  • Allowing both Attorney General enforcement and private lawsuits could provide more than one way to challenge conduct that causes consumer losses.
  • The separate ban on shared AI setting prices could be seen as a precaution against risks that are hard to detect or prove after the fact.

Case against#

  • The ban on shared AI setting prices applies even when the AI does not use competitor data. This could restrict shared pricing tools beyond the bill’s stated concern about nonpublic competitor information.
  • The bill sets large maximum fines, including up to $100 million for each structure developed for distribution, while leaving important questions about evidence and application unresolved.
  • The terms “commercial term” and “common artificial intelligence” may leave businesses uncertain about which tools and activities are covered.
  • The bill does not specify how enforcement would be staffed or funded, or how often the Attorney General would use these powers.