TCAI Bill Guide: Utah’s HB 286, the AI Transparency Act
On February 12, 2026, Utah emerged as a frontrunner in the management of artificial intelligence (AI). The state legislature has implemented laws mandating the disclosure of AI interactions, overseeing the use of AI-driven mental health chatbots, and enhancing protections against AI deepfakes.
This year, lawmakers in Salt Lake City are contemplating HB 286, known as the Artificial Intelligence Transparency Act. This proposed legislation aims to equip developers of large frontier AI models with necessary public safety and child protection plans.
Key Provisions of the Bill
HB 286 specifically targets developers of frontier AI models, which are characterized as the largest and most powerful AI systems. Importantly, this bill exempts smaller models and universities engaged in research involving frontier models.
Public Safety Plan for Catastrophic Risks
The Act mandates that developers of large frontier models establish and publicize a comprehensive public safety plan. This plan must encompass the following elements:
- Adherence to national standards and industry best practices;
- Identification and assessment of thresholds for potential catastrophic risks;
- Application of mitigations to address these risks;
- Utilization of third parties for risk assessments;
- Regular updates to the safety plan;
- Implementation of cybersecurity practices;
- Identification and response to critical safety incidents;
- Establishment of internal governance practices for compliance with the Act.
The term “catastrophic risk” refers to scenarios where a frontier model may facilitate the creation of weapons, engage in cyberattacks, or violate Utah law.
Child Protection Plan
Additionally, the Act requires developers to implement and publish a child protection plan. This plan must include:
- Incorporation of national standards and industry best practices;
- Assessment of potential child safety risks;
- Mitigation strategies for identified risks;
- Third-party evaluations of safety measures;
- Regular updates to the child protection plan;
- Identification and response protocols for child safety incidents;
- Internal governance practices to ensure adherence to these protocols.
The term “child safety risk” refers to any risk that a developer’s model might engage in harmful behavior when interacting with minors.
Transparency and Reporting
Under the Act, developers are prohibited from making materially false or misleading statements regarding their activities, management, or compliance. However, they may redact information in public documents to protect trade secrets or comply with legal requirements, provided they justify these actions.
The Utah Office of Artificial Intelligence Policy may facilitate a reporting mechanism for safety incidents. Developers must report incidents within 15 days and critical incidents within 24 hours. Furthermore, quarterly assessments of catastrophic risks must be submitted to the Office.
Whistleblower Protections
The Act mandates that developers implement internal processes for employees to report safety threats anonymously. Retaliation against whistleblowers is strictly prohibited, with potential penalties for violators.
Enforcement and Penalties
The enforcement of the Act will be overseen by the Utah Attorney General’s Office. Violations can result in civil penalties of up to $1 million for the first offense and up to $3 million for subsequent violations.
Effective Date
The provisions of the Act will take effect on May 6, 2026.
Conclusion
In a recent hearing, Rep. Doug Fiefia emphasized the importance of this legislation, stating, “This bill is not about banning AI; it’s about ensuring safety.” He highlighted the potential risks associated with powerful AI systems, advocating for transparency, risk management, and whistleblower protections.
Joseph Gordon-Levitt, who also testified in favor of the bill, expressed concern for future generations, urging lawmakers to consider the implications of unchecked AI development.
As Utah steps forward to lead in AI regulation, the passage of HB 286 could set a precedent for future legislation aimed at safeguarding public welfare and child safety in the realm of artificial intelligence.
