Understanding California's Latest GenAI Legislation: Implications for AI Developers and AI Consumers
Oct 9, 2024
California Governor Gavin Newsom recently signed several bills that redefine the state's approach to regulating generative artificial intelligence (GenAI). These bills have significant implications for AI developers, companies integrating AI, and consumers alike.
Although these laws are state-specific, they are likely to have a broader impact due to California's influence on technology and policy. As home to tech giants like Google, OpenAI, and Meta, California often sets industry standards that shape global practices.
This article breaks down the key details of each of these bills, who they impact, and what implications they may have on the future of AI.
Bills Affecting AI Developers
AB 2013 - Transparency in AI Training Data
Requires all developers of generative artificial intelligence (GenAI) systems or services released on or after January 1, 2022, regardless of the size of the company or project, to publicly disclose detailed information about the data used to train their AI models by January 1, 2026.
Developers must post comprehensive documentation on their websites, including:
Summaries of datasets used.
Data sources and owners.
Types of data and relevance to the AI's purpose.
Details on copyright, trademark, or patent protections.
Information on personal or consumer data included.
Use of synthetic data.
However, there are specific exemptions outlined in the bill where these requirements do not apply:
Security and Integrity Purposes: GenAI systems developed solely for maintaining the security and integrity of information systems.
Aircraft Operation: Systems used for operating aircraft in the national airspace system.
National Security or Defense: AI developed exclusively for national security or defense purposes and available only to the federal government or its contractors.
Key Takeaways:
If you are a developer or part of a project that has released a generative AI system or service in California after January 1, 2022, you are required to adhere to AB 2013's transparency mandates unless your system falls under the specific exemptions mentioned above. While AB 2013 aims to increase transparency and ethical practices in AI development, many companies are hesitant to disclose their training data due to competitive and legal concerns. This reluctance highlights the tension between regulatory efforts to promote transparency and industry priorities to protect proprietary information and minimize legal risks.
Bill Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB2013
SB 942 - Provenance Disclosure and Identification Tools
Mandates that providers of GenAI systems include identifiable disclosures in AI-generated content and offer free AI detection tools to verify such content.
Providers must:
Include manifest and latent disclosures in AI-generated content.
Provide a free, publicly accessible AI detection tool.
Revoke licenses from third parties who remove required disclosures.
Manifest disclosures are clear, visible labels or audible statements directly indicating that the content is AI-generated, making it understandable to the average user. Latent disclosures are metadata embedded within the content that provides details such as the AI provider’s name, version, and creation date, detectable through AI detection tools. These disclosures are aimed to help users better identify AI-generated material, reduce the risk of deception, and promote trust in the use of generative AI.
Key Takeaways:
SB 942 requires that generative AI systems must embed both manifest and latent disclosures directly into their output to ensure transparency. This means that any content produced by the AI must include visible labels and metadata indicating it was AI-generated. While these disclosures should ideally remain intact if someone copies or shares the content, the legislation focuses on the responsibility of the AI provider to include them at the point of generation. Once a user manually modifies or transfers the content, retaining these disclosures may be less enforceable, depending on the content type.
Bill Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240SB942
Bills Affecting Companies Integrating AI
AB 1008 - AI and the California Consumer Privacy Act (CCPA)
Amends the California Consumer Privacy Act (CCPA)—a state statute intended to enhance privacy rights and consumer protection—to clarify that personal information includes data in physical, digital, and abstract formats, including AI systems capable of outputting personal information.
The bill expands definitions to specifically include:
Personal information across all formats.
Sensitive personal information like neural data.
Obligations for businesses to comply with updated data standards.
This means AI developers must treat all personal information—including data in physical, digital, or abstract formats processed or generated by AI systems—in full accordance with the California Consumer Privacy Act (CCPA). This involves updating privacy policies to transparently disclose how personal data is collected, used, and shared; obtaining explicit consent from individuals before collecting their data; providing clear mechanisms for consumers to exercise their rights to access, delete, or opt out of the sale of their personal information; and implementing strong security measures to protect this data.
Developers must also ensure that their AI systems do not inadvertently output or misuse personal information and that any training data is sourced and handled legally and ethically.
Key Takeaways:
AI developers must align their practices with AB 1008 by ensuring that all personal information—including data processed or output by AI systems—is handled in compliance with the CCPA. This involves updating privacy policies for transparency, obtaining proper consent for data collection, respecting consumer rights to access, delete, or opt-out of the sale of their personal information, implementing robust security measures to protect personal data, and ensuring that AI systems do not inadvertently misuse or disclose personal information.
Bill Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB1008
AB 2602 - Consent and Digital Replicas
Ensures contract provisions involving the use of digital replicas of an individual's voice or likeness are enforceable only with explicit and informed consent.
According to the bill, contracts are unenforceable if they:
Allow use of digital replicas without specific intended uses.
Were agreed upon without the individual's legal representation.
Enable replacement of the individual's work with a digital replica without proper consent.
Key Takeaways:
To comply with AB 2602, AI developers and companies must ensure that contracts involving the use of digital replicas of an individual's voice or likeness include a clear and specific description of the intended uses of the replica. Additionally, they must confirm that individuals have legal representation during contract negotiations to validate consent. Failure to meet these requirements will render the contract provisions unenforceable, thereby protecting individuals from unauthorized or unethical use of their digital identity.
Bill Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB2602
AB 3030 - GenAI in Healthcare Communications
AB 3030 requires healthcare providers using generative AI (GenAI) for patient communications to include clear disclaimers and provide contact information for human providers. Specifically, AI-generated written, audio, and video communications must prominently state their AI origin:
written materials should have disclaimers at the beginning,
audio interactions need verbal disclaimers at both the start and end,
and video content must display disclaimers throughout.
Additionally, patients must be given instructions on how to contact a human healthcare provider for further assistance. An exemption applies if a licensed human provider reviews and personally delivers the AI-generated communication.
Key Takeaways:
Healthcare providers need to update their communication protocols to ensure that AI-generated patient materials include the required disclaimers and human contact information. Providers should also establish a process for having licensed human professionals review AI communications when necessary to bypass the disclaimer requirement. Staff must be trained on these updated procedures, and regular compliance audits should be conducted to verify adherence to AB 3030, ensuring transparency and patient trust.
Bill Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB3030
SB 896 - State Use and Risk Analysis of GenAI
SB 896, the Generative Artificial Intelligence Accountability Act, establishes requirements for California state agencies using generative AI (GenAI) to ensure transparency, privacy, equity, and accountability.
State agencies must:
Conduct annual risk analyses on GenAI threats and report findings to the Legislature.
Include disclaimers in AI-generated communications, ensuring they are clear and prominent, and provide contact information for a human representative.
Promote responsible AI usage by safeguarding privacy, minimizing bias, and ensuring transparency in their AI systems.
Engage in workforce development to train staff for AI roles, focusing on ethics and privacy.
Key Takeaways:
SB 896 aims to build public trust in AI technology, which is essential for the sustainable growth of the AI ecosystem. By enforcing transparency, privacy, and accountability in AI-generated communications used by state agencies, the law sets a precedent for responsible AI use. This helps ensure that AI is seen as a reliable tool, which can boost consumer confidence and encourage wider adoption, ultimately supporting the industry's growth and acceptance.
Bill Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240SB896
Bills Affecting Consumers Using AI
AB 2355 & AB 2655 - AI in Political Advertising
These bills introduce regulations on the use of AI and deepfake content in political advertising and election-related content to promote transparency and integrity.
Specifically:
AB 2355: Requires political ads using AI-generated or altered content to include clear disclosures stating they were generated using AI.
AB 2655: Mandates online platforms to block or label deceptive election-related content and provides mechanisms for reporting such content.
Key Takeaways:
These bills set a crucial precedent for protecting the integrity of democratic processes as generative AI (genAI) capabilities continue to advance, making it increasingly difficult to distinguish fact from fiction. By requiring clear disclosures for AI-generated political ads (AB 2355) and mandating platforms to label or remove misleading election content (AB 2655), these laws aim to curb the influence of manipulated media and deepfakes. This regulatory framework is essential for ensuring transparency in political communication and maintaining voter trust in an era where AI-generated content can easily deceive.
AB 2355 Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB2355
AB 2655 Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB2655
AB 1836 - Digital Replicas of Deceased Individuals
AB 1836 prohibits the unauthorized commercial use of a deceased person's voice or likeness through digital replicas without prior consent from the rights holder. This law targets individuals and entities that produce, distribute, or make available such digital replicas for commercial purposes without authorization. By imposing significant damages on violators, AB 1836 aims to protect the legacy and rights of deceased individuals and prevent exploitation through AI-generated content.
Key Takeaways:
As AI technology continues to advance, replicating anyone's voice or likeness will become increasingly accessible, raising concerns about misuse and exploitation. AB 1836 is a step forward in addressing these ethical challenges by establishing legal requirements for the commercial use of digital replicas of deceased individuals. By mandating explicit consent from rights holders and imposing significant penalties for unauthorized use, the bill helps protect personal identities from exploitation and sets a precedent for responsible use of generative AI. This law contributes to a more ethical framework for AI development, ensuring that as replication technologies evolve, individuals' rights and legacies are safeguarded.
Bill Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB1836
SB 926 & SB 981 - Protection from AI-Generated Explicit Content
SB 926 and SB 981 are legislative measures aimed at curbing the harmful and unauthorized use of generative AI to create explicit content without consent. These laws address the misuse of deepfake technologies and hold creators and platforms accountable for their role in protecting individuals' privacy and digital identity.
SB 926 criminalizes the intentional creation and distribution of realistic, AI-generated intimate images of individuals without their consent. This includes non-consensual deepfakes depicting a person in a sexually explicit manner, with penalties to deter such misuse.
SB 981 requires social media platforms to provide mechanisms for users to report instances of sexually explicit digital identity theft. Platforms must investigate and remove such content swiftly, providing victims with avenues to seek prompt redress.
Key Takeaways:
As generative AI becomes increasingly accessible, the potential for misuse—such as creating non-consensual explicit content—grows significantly. SB 926 and SB 981 establish crucial protections by criminalizing unauthorized AI-generated intimate content and holding platforms accountable for prompt removal of harmful material. These measures underscore the importance of ethical AI use, emphasizing privacy, consent, and the need for responsible technology deployment to safeguard individuals from emotional and reputational harm as genAI continues to advance.
SB 926 Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240SB926
SB 981 Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240SB981
AB 1831 & SB 1381 - Expanding CSAM Statutes
AB 1831 and SB 1381 are legislative measures that expand the definition of child sexual assault material (CSAM) to include content generated or altered by artificial intelligence (AI) or digital means depicting minors. These bills make it illegal to produce, distribute, or possess AI-generated CSAM, with severe penalties including felony charges and imprisonment. The legislation addresses the emerging threat posed by advanced technologies that can create realistic but artificial depictions of minors in illegal content, closing legal loopholes that previously existed.
Key Takeaways:
AB 1831 and SB 1381 are vital in upholding ethical standards and protecting vulnerable populations as AI technology becomes more sophisticated and accessible. While AI developers are not directly targeted by AB 1831 and SB 1381, they should be proactive in implementing measures to prevent their technologies from being used to produce AI-generated CSAM. By doing so, they help combat the exploitation of minors, align with ethical standards, and reduce potential legal and reputational risks.
AB 1831 Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240AB1831
SB 1381 Text:
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240SB1381
Implications for the Future of AI
California's GenAI legislation sets a new standard for responsible AI governance. By emphasizing transparency, privacy, and ethical use, these laws aim to mitigate risks without stifling innovation. They ensure that AI development aligns with societal values, protecting individuals and fostering trust in AI technologies. As AI continues to integrate into daily life, these types of regulatory frameworks ensure that AI remains a positive force of technological advancement.
To stay at the forefront of ethical AI development and leverage AI’s full potential, understanding the implications of regulatory changes like California’s GenAI laws is crucial. At Sahara AI, we are committed to building a decentralized, transparent, and privacy-first AI ecosystem that ensures security, equity, and accessibility for all users. Join our waitlist to stay up to date on the latest advancements and be part of the movement shaping the future of ethical AI.
