SB-53 Assembly Privacy analysis 2025-09-05

Citation

California State Assembly, Committee on Privacy and Consumer Protection (Rebecca Bauer-Kahan, Chair), analysis of SB 53 (Wiener) as amended September 5, 2025, for the hearing of September 11, 2025. https://apcp.assembly.ca.gov/system/files/2025-09/sb-53-wiener-apcp-analysis-9.11.pdf. Wiki copy: raw/legislative/SB-53 Assembly Privacy analysis 2025-09-05.md. Read 2026-10-15.

They are lines 10-20, 121-135, 166, 169, 179, 183, 233-237, 285-286 and 340 of a full transcription. [...] marks the cuts.

Bottom line

• This is the last committee analysis of SB 53, on the version the Governor signed. Its CalCompute text repeats the July 8 analysis.
• The bill’s subject changed. In July it was “CalCompute: foundation models: whistleblowers”. In September it is “Artificial intelligence models: large developers”.
• Item 27 opens with “Upon appropriation”, a short form of subd. (k).
• The University of California is a registered supporter of SB 53 in this version too.
• It repeats the systemic risk passage on “the high costs of developing AI systems”.

Background

The committee heard the bill on September 11, 2025. The analysis covers the expanded bill: the frontier AI transparency rules plus CalCompute. The date in the file name, September 5, is the amendment date.

Key claims

• Item 27.

  27) Upon appropriation, establishes within GovOps a consortium to develop a framework for thecreation of a public cloud computing cluster to be known as “CalCompute” that advances thedevelopment and deployment of artificial intelligence that is safe, ethical, equitable, andsustainable by doing, at a minimum, both of the following:

• Items 28 to 32 restate subd. (c), (d), (f)(1), (g) and (j). Item 30 says the Report goes to the Legislature “as specified” and does not list the seven elements.

• Comments, “CalCompute”. The text repeats the July 8 analysis with small word changes, for example “like SB 1047” in place of “similarly to SB 1047”. It again states the “resource and research gap” and “The consortium must prioritize locating CalCompute within the University of California system.”

• International AI Safety Report. “The recently released International AI Safety Report, developed by nearly 100 internationally recognized experts from 30 countries led by Turing Award winner Yoshua Bengio, sets forth three general risk categories associated with frontier models: malicious use, malfunctions, and systemic risk.”

• Systemic risks, from that report. “Such risks include labor market disruption, global AI research and developmentconcentration, market concentration, single points of failure, environmental risks, privacyrisks, and copyright infringement.”

• Systemic risks, the committee’s comments.

  Systemic risks. Due to the high costs of developing AI systems, a small number of largetechnology companies dominate the frontier model market, compounding many of the risksdescribed above. Widespread use of a few frontier models can make critical sectors such ashealthcare and finance vulnerable to systemic failures if a model has flaws, vulnerabilities, bugs, or biases.[^33]

• Registered Support: University of California. The list of supporters under “REGISTERED SUPPORT / OPPOSITION” includes “University of California”. UC is a registered supporter of SB 53.

Evidence and method

A committee staff summary of the bill. No new data on CalCompute.

Limits

• The excerpt’s numbered list does not show subd. (e) (labor laws), (h), (i) or (k) as separate items. The full analysis may list them elsewhere.

• The transcription lost many spaces (“thecreation”, “andsustainable”).

• “Upon appropriation” shortens subd. (k).

• The excerpt shows “University of California” in the support list but not the rest of the list.