SB 53 and the TDTA: How California is regulating AI transparency

When Gov. Gavin Newsom, D-Calif., signed California Senate Bill 53, the Transparency in Frontier Artificial Intelligence Act, in September 2025, it quickly became the centerpiece of the state's broader push to regulate AI. 

SB 53 was not the only AI transparency law California had enacted, though. A year earlier, Newsom signed Assembly Bill 2013, the Generative Artificial Intelligence Training Data Transparency Act. Both laws took effect 1 Jan. 2026. Both are transparency statutes. 

But where SB 53 asks a small number of frontier developers what could go wrong with their models, the TDTA asks a wider range of developers what went into building them. Together, the two laws represent complementary pieces of California's approach to AI regulation: while SB 53 focuses on the safety and oversight of frontier AI models, the TDTA focuses on the data used to train them.

The difference in scope between SB 53 and the TDTA is the most important distinction for organizations trying to determine their compliance obligations. SB 53 applies exclusively to "frontier developers" and "large frontier developers," defined by a 10 to the power 26 floating-point operations per second computational threshold and, for large frontier developers, annual gross revenues exceeding USD500 million. In practice, this limits SB 53's reach to a handful of companies training the largest foundation models.

The TDTA, on the other hand, applies much more broadly. It applies to any "developer" that "designs, codes, produces, or substantially modifies an artificial intelligence system or service for use by members of the public." The statute defines "substantially modifies" and "substantial modification" as "a new version, new release, or other update to a generative artificial intelligence system or service that materially changes its functionality or performance, including the results of retraining or fine tuning."