On 29 September 2025 California’s governor signed SB 53, the Transparency in Frontier Artificial Intelligence Act. The statute places frontier AI model development within a “transparency, accountability and reporting” framework, and is a significant state-level legislative move on frontier model governance in the United States.
Four pillars of the Act
Published safety frameworks. Large frontier AI developers must publish a framework on their website explaining how they incorporate national, international and industry standards into their frontier AI systems. The requirement is not a particular technical solution but that the governance logic be externally reviewable.
Critical safety incident reporting. Companies must report critical safety incidents to the state emergency services body, moving frontier model risk from an internal corporate matter to something that must be disclosed to a public authority.
Whistleblower protection. Protection for insiders who disclose serious health and safety risks lowers the barrier to internal information reaching the outside.
Civil penalty authority. The governor may impose civil penalties for violations, giving the transparency requirements an enforcement mechanism.
Tension with the federal path
At the federal level, US AI policy in 2025 has leaned towards reducing fragmentation and emphasising a single rule. See our notes on the US executive order on a national AI policy framework and the US AI model review executive order.
There is tension between state law and the federal direction: the federal approach prefers one rule to reduce compliance burden, while state law prefers to build accountability where the federal government has not. The tension is particularly acute in AI, because frontier developers are often concentrated in the state, giving state law real effect.
What it means for AI companies
First, public disclosure becomes part of compliance. Safety frameworks, incident reporting and risk disclosure mean companies need a consistent line between internal governance and external communication.
Second, an incident reporting mechanism is needed. Companies should define what counts as a critical safety incident, the internal escalation path and the external reporting process.
Third, the cost of stacking state and federal rules. Companies operating in the US must track both the federal direction and legislative developments in key states such as California, and keep compliance design adjustable. See our multi-jurisdiction service and our note on the US national AI framework.
SB 53 signals that in frontier model governance, legislators’ focus has shifted from model capability to developers’ transparency and accountability. That raises the visibility requirements on corporate governance.
This article is based on California’s SB 53, the Transparency in Frontier Artificial Intelligence Act, and on public information. It is for general reference only and does not constitute legal advice.