|
REFERENCE TITLE: artificial intelligence business; attorney general |
|
State of Arizona House of Representatives Fifty-seventh Legislature Second Regular Session 2026
|
|
HB 4098 |
|
|
|
Introduced by Representative Travers
|
AN ACT
amending title 44, chapter 9, arizona revised statutes, by adding article 27; relating to commerce.
(TEXT OF BILL BEGINS ON NEXT PAGE)
Be it enacted by the Legislature of the State of Arizona:
Section 1. Title 44, chapter 9, Arizona Revised Statutes, is amended by adding article 27, to read:
ARTICLE 27. ARTIFICIAL INTELLIGENCE BUSINESS
44-1383. Artificial intelligence business; risk assessment; report; posting; quarterly audit; violation; civil penalty; definitions
A. An artificial intelligence business shall conduct a comprehensive risk assessment for each of the business's high risk AI systems before using or selling a high risk AI system.
B. An artificial intelligence business shall evaluate a high risk AI system for all of the following:
1. The potential for discrimination or bias.
2. Any safety risks to children and vulnerable populations.
3. Any privacy and data protection risks.
4. assessing catastrophic risks.
C. An artificial intelligence business shall submit a transparency report to the attorney general before using or selling a high risk AI system. THe transparency report must include all of the following:
1. The purposes and operation of the artificial intelligence system.
2. The results from any internal risk assessment evaluations.
3. The measures that were implemented to mitigate bias and harmful outcomes.
D. The transparency reports shall be publicly posted on the artificial intelligence business's website.
E. An artificial intelligence business shall implement bias mitigation measures that include any technical or procedural methods that were implemented to reduce discriminatory outcomes.
F. An artificial intelligence business shall conduct a quarterly audit to identify and correct any instances of discrimination, harmful outputs or unsafe behavior.
G. If a high risk AI system may be accessed by a child, the artificial intelligence business shall include parental monitoring tools, content filtering and mechanisms to report harmful content.
H. The attorney general shall issue a certification before an ARTIFICIAL intelligence business enters the stream of commerce.
I. The attorney general shall enforce this section and review the transparency reports for compliance with this section. The attorney general may impose a civil penalty of not more than $50,000 per violation for a violation of this section. Each day a violation occurs constitutes a separate violation.
J. An artificial intelligence business that implements recognized standards for and complies with the children's online privacy protection act of 1998 (P.L. 105-277) advisory services may receive a reduced civil penalty for inadvertent violations.
K. For the purposes of this section:
1. "Artificial intelligence business" means an entity that creates, uses, sells or provides access to artificial intelligence systems that interact with the public and that uses a frontier model of artificial intelligence that is owned or operated by a large developer.
2. "frontier model" means an artificial intelligence model that is trained by either of the following:
(a) using greater than 10.26 computational operations to compute costs of more than $100,000,000.
(b) applying knowledge distillation to a frontier model.
3. "High risk AI system" means an artificial intelligence system that:
(a) May impact children, health care, employment, public safety or education.
(b) Is likely to cause discrimination, bias or significant privacy violations.
4. "Large developer":
(a) Means a person or entity that has trained one or more frontier models and has spent more than $5,000,000 in developing the frontier model and more than $100,000,000 in training the frontier model.
(b) Does not include an accredited college or a university that engages in academic research involving frontier models.
(c) Creates a FORESEEABLE and material risk that a large developer's development, storage or deployment of a foundation model will result in the death of or serious injury to more than one hundred people by an incident of either of the following:
(i) creating and releasing a chemical, biological, radiological or nuclear weapon.
(ii) through the use of A FOUNDATION model, engaging in conduct that is performed with limited human intervention and, if the conduct was committed by an individual, would constitute a crime that requires intent, recklessness or gross negligence or the SOLICITATION or aiding and abetting of a crime.