Over the past year, lawmakers in California have moved forward with a series of artificial intelligence (“AI”)-focused laws aimed at increasing transparency, accountability, and safety, as AI becomes more integrated into key industries around the country and world. The legislation that the state has spearheaded addresses a wide range of issues – from data disclosure and content identification to consumer privacy and education – and also maintains California’s status as an early-mover when it comes to setting standards for new technology and sustainability practices.
AB 2013, for example, requires developers to disclose the datasets used in training AI models by 2026, which is intended to give regulators and consumers insight into the data that powers AI systems, Morgan Lewis’ Minna Naranjo, Dion Bregman, Andrew Gray, and Leigha Beckman stated in a recent note. Meanwhile, SB 942 mandates that AI-generated content include watermarks, which will presumably help consumers distinguish between AI-created and human-made materials, a measure that lawmakers have said is critical to sectors like media and education.
In a separate legislative effort, AB 3030 aims to ensure that patients are informed when AI tools are involved in their care, while AB 2876 is pushing to get AI literacy included in public school curricula in order to prepare students for an AI-driven future.
The legislation in California is at various stages of implementation, with some laws, like AB 2013 and SB 942, for example, set to take effect by 2026, while others like SB 1047 – an ambitious piece of legislation that proposed strict safety measures for large-scale AI models – did not make it past the governor’s desk. TFL provides a breakdown of seven critical AI-centric laws here …
Defining AI (AB 2885)
Introduced: February 15, 2024 by Assemblymember Rebecca Bauer-Kahan
Snapshot: AB 2885 formally defines AI under California state law, providing a consistent legal framework for AI regulation.
Key Provisions: AB 2885 defines artificial intelligence as an “engineered or machine-based system that varies in its level of autonomy” and that can generate outputs based on input data. The law creates a uniform legal definition, helping regulators and developers align their understanding of AI across various applications. It will become effective on January 1, 2025.
Potential Implications: By standardizing the definition of AI, the bill aims to offer clarity for future regulations and helps ensure that the term is applied consistently across different industries.
Artificial Intelligence and Training Data Transparency (AB 2013)
Introduced: January 31, 2024 by Assemblymember Jacqui Irwin
Snapshot: This bill targets the transparency of datasets used in training generative AI systems. It is a key part of California’s efforts to make AI development more open and accountable, lawmakers said.
Key Provisions: AB 2013 mandates that developers of generative AI models reveal specific details about the datasets they use in the development and training of their AI tools. The California Department of Technology will monitor compliance with the law, requiring that the necessary disclosures be made public by January 2026. By demanding transparency, this law seeks to address concerns over the biases and potential risks that might stem from undisclosed or poorly sourced data, according to the bill.
Potential Implications: If enacted, this act would provide a significant step toward building more trust in AI systems. Consumers and regulators will have greater insight into the foundation of these models, which experts agree is crucial for ensuring that AI is fair and unbiased. Furthermore, increased transparency may also lead to a broader adoption of ethical AI standards.
California AI Transparency Act (SB 942)
Introduced: January 17, 2024 by Sen. Josh Becker
Snapshot: SB 942 focuses on making sure AI-generated content is clearly identifiable, while also supporting the public’s ability to distinguish between AI and human-generated content.
Key Provisions: This law, which Newsom signed in September, enforces the use of visible watermarks on all AI-generated content to help people recognize when they are interacting with material created by AI. Developers are also required to create detection tools, which must be freely available to the public, ensuring transparency. Furthermore, any third parties using these AI systems must keep the watermarking intact, as stipulated by their contracts. This regulation ensures consistency in how AI is identified across different platforms and sectors.
Potential Implications: This law enhances transparency in AI-driven industries like media, education, and communications, where the public often engages with AI-generated material unknowingly. By clearly labeling AI content, it helps prevent misuse, especially in areas like election advertising or educational content.
Generative Artificial Intelligence Accountability Act (SB 896)
Introduced: January 3, 2024 by Sen. Bill Dodd
Snapshot: This legislation, signed into law at the end of September, brings AI oversight into state agencies, ensuring that Californians are aware when they are interacting with AI systems in government settings.
Key Provisions: Under SB 896, state agencies are required to submit reports evaluating both the risks and benefits of using generative AI tools in their operations. Additionally, agencies must disclose to the public when AI systems are used in any direct interactions with citizens, whether online or in-person. This regulation marks a shift toward greater governmental transparency regarding the deployment of AI technologies.
Potential Implications: This bill tackles AI accountability by making the state’s use of AI tools more transparent. It also provides citizens with clear information about when they are engaging with AI, which could help foster public trust in the government’s use of technology.
California Consumer Privacy Act Amendment (AB 1008)
Introduced: July 2, 2024 by Assemblymember Rebecca Bauer-Kahan
Snapshot: This amendment ensures that AI systems handling personal information fall under the scope of California’s already stringent privacy laws.
Key Provisions: AB 1008 expands the scope of the California Consumer Privacy Act to include AI systems that process personal data. This means that developers using AI to generate personal information outputs must now adhere to the strict privacy guidelines that apply to other data processors in California. The law addresses the growing concerns over how AI might mishandle sensitive personal information.
Potential Implications: By extending privacy protections to AI systems, this amendment seeks to strengthen consumer trust and ensure that Californians’ personal data remains safeguarded in an AI-driven environment. This legislation is a response to the increasing integration of AI in industries that process personal data, from healthcare to finance.
AI in Healthcare Disclosure (AB 3030)
Introduced: February 16, 2024 by Assemblymember Lisa Calderon
Snapshot: AB 3030 brings transparency into healthcare by requiring that providers inform patients when AI tools are involved in their care.
Key Provisions: This legislation mandates that healthcare providers must disclose whenever they use AI tools in communications or treatment plans with patients. The aim is to make patients fully aware of when AI technology is involved in their care, ensuring informed consent. The law recognizes that AI tools are increasingly integrated into healthcare, whether in diagnostics, patient management, or administrative processes.
Potential Implications: AB 3030 endeavors to improve trust in healthcare by ensuring transparency in AI’s role in patient care. It emphasizes patient autonomy, giving people the information they need to make informed decisions about their health.
AI Literacy in Schools (AB 2876)
Introduced: February 15, 2024 by Assemblymember Marc Berman
Snapshot: AB 2876 is designed to ensure that AI literacy becomes an integral part of California’s public school curriculum, preparing students for the future.
Key Provisions: The legislation directs the state superintendent to create working groups to assess how AI can be incorporated into school curriculums. The goal is to develop standards for AI literacy that can be integrated across K-12 schools. The law also requires that safety guidelines be put in place to ensure that students and educators use AI responsibly.
Potential Implications: This bill positions California to lead the way in AI education. By introducing AI literacy to students at an early age, it aims to prepare future generations to work and thrive in an increasingly AI-driven world.
Telemarketing and AI-Generated Voices (AB 2905)
Introduced: February 15, 2024 by Assemblymembers Evan Low and Jim Patterson
Snapshot: AB 2905 focuses on telemarketing and requires that AI-generated voices be disclosed during robocalls.
Key Provisions: This bill addresses the use of synthetic voices in telemarketing, requiring telemarketers to clearly state when AI-generated voices are used during phone calls. This is intended to prevent confusion or deception, particularly in cases where AI-generated voices might mimic human ones, according to the bill.
Potential Implications: This law is part of a broader effort to increase transparency and accountability in AI applications, especially in commercial and political contexts. It aims to protect consumers from manipulative or misleading communications.
Share
