California has enacted the Transparency in Frontier Artificial Intelligence Act (TFAIA) through Senate Bill 53, effective January 1, 2026. The law targets large developers of “frontier models,” defined as models requiring more than 10^26 operations during training, fine-tuning, or modification, and whose annual gross revenue exceeds $500 million.
Under TFAIA, developers must publish documentation on training data and implement a frontier AI framework that incorporates national and international standards and industry best practices. They are required to submit summaries of catastrophic risk assessments to the Office of Emergency Services (OES) and report critical safety incidents through mechanisms established by OES. Violations may result in civil penalties of up to $1 million per violation, particularly for failures that could lead to death or serious injury of more than 50 people or property damage exceeding $1 billion.
Catastrophic risk is defined as a foreseeable and material risk that AI development or deployment could cause mass harm, including aiding in the creation of weapons, conducting cyberattacks or criminal acts without oversight, or evading developer control. The statute excludes lawful federal activity, harm caused by unrelated software, and publicly accessible information.
The law also introduces whistleblower protections for employees involved in risk assessment, safety management, or incident response. Employers must provide annual notices of these rights, allow anonymous internal reporting, and prohibit retaliation. Employees may report concerns to authorities and seek damages, injunctive relief, and attorney’s fees for violations. Employers must give monthly updates on investigations and share summaries with company officers quarterly, except when implicated.
Additionally, the Act preempts local regulations adopted after January 1, 2025, related to catastrophic risk management and requires compliance with notice and acknowledgment procedures for all employees, including remote workers.