Governor Newsom Vetoes AI Safety Bill While Signing Multiple AI-Related Laws
Sacramento, California – The fate of Senate Bill 1047, a highly debated AI safety bill, has finally been determined. California Governor Gavin Newsom vetoed the bill after weeks of deliberation. SB 1047 was designed to require developers of advanced AI models and entities providing necessary computational resources to establish robust safeguards and policies to mitigate critical risks associated with AI technology. Despite recognising the bill's good intentions, Governor Newsom stated that it might not be “the best approach to protecting the public from real threats posed by the technology.”
In the midst of the intense discussions surrounding SB 1047, Governor Newsom has approved 17 other bills addressing various aspects of generative AI (GenAI). These include regulations on deepfakes, AI watermarking, child safety, performers’ AI rights, and election misinformation. These legislative moves indicate an active stance on the regulation and ethical oversight of AI technologies in the state.
Among the newly signed legislation, Assembly Bill 2013 stands out. This law necessitates that by January 1, 2026, developers of GenAI systems or services released on or after January 1, 2022, must publish a “high-level summary” of the datasets used to train these systems. The summary must be made available to the public in California and should include details such as:
- The sources or owners of the datasets.
- The approximate number of data points and a description of their types.
- Any intellectual property protection associated with the datasets.
- Whether the datasets were purchased or licensed.
- Inclusion of personal information or consumer aggregate data as defined under the California Consumer Privacy Act (CCPA).
- Any cleaning, processing, or modification carried out on the datasets.
- The period during which the data was collected or used.
- The use of synthetic data generation in the development process.
Importantly, AB 2013 exempts GenAI systems dedicated solely to data security and integrity, or those utilised by federal entities for national defence purposes, among other exclusions.
The introduction of AB 2013 into law places California at the forefront of regulating GenAI training data sources. Disclosure of data sources has long been a sensitive issue for GenAI developers due to competitive and business considerations. Additionally, this step coincides with ongoing copyright and privacy litigations where major GenAI developers have been accused of employing copyrighted materials and personal data without proper authorization.
Other significant AI-related legislations signed by Governor Newsom include:
AB 1008: This law clarifies the definition of “personal information” under the CCPA, extending it to encompass various formats, including physical, digital, and abstract digital formats, such as AI systems capable of producing personal information.
SB 1120: This law mandates that health care service plans or disability insurers utilising AI for coverage decisions must ensure that the AI tool bases its determinations on specified information and applies these considerations fairly and equitably.
AB 3030: This bill requires health facilities, clinics, physicians' offices, or group practices using GenAI for patient communications to include a disclaimer. This disclaimer must inform patients that GenAI was used and provide instructions on how to contact a human healthcare provider.
By approving these bills, Governor Newsom has highlighted the need for a balanced and considered approach to AI regulation. The new laws aim to enhance transparency, protect personal information, and ensure ethical AI practices across various sectors. This legislative activity underscores California’s proactive stance in managing the integration of AI technology within society.
Source: Noah Wire Services