California Governor Gavin Newsom recently signed into law a wave of legislation – totaling 19 laws – addressing the opportunities and risks of AI and placing California at the forefront of AI regulation in the United States. From election integrity to performer rights and healthcare transparency, the state has enacted measures aimed at managing potential negative impacts of the AI boom. At the same time, Governor Newsom vetoed SB 1047, the most comprehensive bill on his desk, signaling his interest in balancing the need for regulation to promote the safe deployment of AI with an interest in fostering growth in this important new sector of the California tech economy.
Enacted AI Legislation:
These new laws, summarized and grouped by subject matter below, are a patchwork of regulation that is likely to grow and change over the coming months and years.
Election Integrity and Disinformation in the Age of AI: Recognizing risks posed by AI-generated disinformation in elections, California enacted three laws to address this emerging issue:
- AB 2655 (Defending Democracy from Deepfake Deception Act of 2024): Requires large online platforms to set up a reporting system and then to label or take down “materially deceptive content” AI-generated content 120 days before and after an election, and continue to label or takedown content for 60 days after an election where deceptive content calls into question the validity of the electoral process. The law also provides for enforcement options for the State AG and providing enforcement mechanisms for candidates and officials. Exemptions apply to certain media and satire.
- AB 2355: Mandates that audio, video, imaged based political ads generated or altered by AI carry a disclaimer, and outlining specific formatting requirements based on the advertising medium The Fair Political Practices Commission will enforce compliance.
- AB 2839[i]: Prohibits the malicious distribution of deceptive election materials, explicitly including AI-generated content, and expands the prohibition period consistent with AB 2665’s 120- and 60-day periods. The law provides a private right of action for injunctive or equitable relief and fee-shifting for successful plaintiffs.
Transparency in AI Development and Use: California passed several laws addressing transparency in the development and use of AI:
- SB 942 (California AI Transparency Act): Requires developers of widely-used generative AI systems to provide a free AI detection tool for users and mandates both visible and hidden “watermark” disclosures on AI-generated content to improve transparency.
- AB 2013 (Artificial Intelligence Training Data Transparency Act): Imposes transparency requirements on generative AI developers to disclose the datasets used for training, including the data’s origins and whether personal or copyrighted information is involved.
- AB 2885: Standardizes the definition of "artificial intelligence" across various California statutes, facilitating consistent application in state agencies, social media compliance, and education systems.
Consumer Privacy and AI Transparency: California has amended the California Consumer Privacy Act (CCPA) to close potential loopholes:
- AB 1008: Clarifies that the CCPA definition of “personal information” includes personal information which is within an AI system capable of outputting personal information.
Protections Against AI-Generated Harmful Sexual Content: Recognizing the potential for AI to generate harmful sexual content, California passed the following laws:
- AB 1831 and SB 1381: Clarifies and expands child pornography laws to include AI-generated child sexual abuse material, criminalizing the production, distribution, and possession of such content.
- SB 926: Criminalizes the creation or distribution of non-consensual, AI-generated intimate images, addressing the emotional harm caused by deepfake content.
- SB 981: Requires social media platforms to provide reporting mechanisms for California residents to report non-consensual sexually explicit digital identity theft, also known as deepfake pornography. The law requires online platforms to promptly investigate, report, and remove violative content, generally within 30 days and blocking the content during the pendency of the investigation.
Regulation of AI in Specific Sectors: In addition to the generally applicable laws above, the set of new laws addresses AI applications in specific sectors as well:
Healthcare and Insurance:
- AB 3030: Mandates disclosure of AI in patient communications and ensures patients have direct communication options with human providers.
- SB 1120: Requires health care plans using AI for reviewing patient care to base decisions on individual patient information, have licensed professionals make medical decisions, and ensure proper oversight of AI systems.
Education:
- AB 2876: Mandates the inclusion of AI literacy in California's curriculum frameworks for subjects such as mathematics, science, and history, preparing students for AI's role in everyday life.
- SB 1288: Establishes a working group tasked with guiding the use of AI in public schools, creating guidelines for safe AI use, and developing a model policy for responsible AI integration.
- AB 2602: Regulates the use of “digital replicas,” AI-generated likenesses of performers, requiring explicit terms in a contract describing the intended use of the replica and that performers have legal representation.
- AB 1836: Prohibits the unauthorized commercial use of a deceased performer's digital replica, protecting postmortem publicity rights and requiring consent from their estate.
Telemarketing:
- AB 2905: Requires disclosures for the use of AI-generated voices in telemarketing calls.
Government Agencies:
- SB 896: Regulates the use of GenAI in California's state government, mandating notification of AI interactions in online interfaces or by phone, and risk analyses and reporting related to the benefits and risks of AI, especially as related to critical infrastructure.
Vetoed Legislation: SB 1047
One major piece of legislation that did not make it into law was SB 1047, which aimed to introduce stringent safety requirements for developers of large-scale AI models costing over $100 million to train. Key provisions included:
- Mandatory Safety and Security Protocols: Developers were required to implement and publish safety procedures, with access to be granted to the California Attorney General.
- Emergency Shutdown Capabilities: Developers needed to ensure their AI systems could be promptly shut down in emergencies.
- Risk Assessments: Developers were mandated to conduct pre-deployment risk assessments for potential harms and to review safety protocols annually.
- Whistleblower Protections: Provisions were included to protect employees who anonymously report safety violations.
- Know Your Customer (KYC) Obligations: Operators of computing clusters had to maintain records of customers using significant resources to train AI models.
Governor Newsom vetoed the bill, citing concerns that it applied stringent standards even to low-risk applications of AI, potentially stifling innovation. He also highlighted the need for adaptability and differentiated regulation, particularly when smaller developers could be disproportionately affected by broad mandates. The governor indicated that while public safety is paramount, future legislation should be informed by empirical data and should not unnecessarily hinder AI advancements.
Looking Forward: States Lead the Way in AI Regulation
As California enacts a diverse set of AI regulations, participants in the AI sector should remain vigilant about other emerging state-level legislation. With multiple states expanding their regulatory frameworks the patchwork of AI laws across the United States is becoming increasingly complex, presenting both challenges and opportunities for shaping best practices and compliance postures. Staying informed and adaptable will be crucial for navigating the evolving landscape of AI governance and ensuring alignment with state-specific requirements. It’s important that individuals, businesses, and other entities involved in the deployment of AI seek legal advice about the applicability and impact of such laws.
If you have questions about these or other laws affecting AI development and deployment, contact Beth George, Janet Kim, Sean Quinn, Madeline Cimino, Christine Chong, or other FBD attorneys that advise you.
[i] A majority of the provisions in AB 2839 were temporarily blocked from enforcement in California Federal Court on October 2, 2024 on constitutional grounds. Requirements pertaining to audio-only content remain enforceable.
