California Bill (SB-1047) Regulating AI Making Its Way through Legislation
California Bill (SB-1047) Regulating AI Making Its Way through Legislation
On August 15, the California State Assembly Appropriations Committee passed Senate Bill 1047 with amendments. California Senate Bill 1047, the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, was put forward in response to the growing concerns about the potential risks and harms associated with large-scale AI models.
California joins other states, including Illinois and Colorado, in proposing and enacting legislation regulating the development, deployment and use of AI technologies. In addition, the Federal Government has released an Executive Order guiding the development and use of AI within government agencies. As of yet, there is no comprehensive Federal legislation specifically regulating AI in the United States.
Why SB-1047
The rapid development and deployment of AI systems have outpaced existing regulatory frameworks, creating a pressing need for proactive measures to ensure public safety and prevent unintended consequences. SB-1047 aims to bridge this gap by establishing a foundation for the responsible and accountable development of advanced AI technologies.
SB-1047 focuses on 1) companies that develop and deploy “frontier AI models,” which it defines as characterized by their ability to learn and make autonomous decisions with significant real-world impact, and 2) companies that engage in business activities within the state of California, such as selling products or services, having employees or offices in California, or targeting California residents with their AI models.
The bill places responsibility on developers to assess and mitigate potential harms through rigorous testing and independent third-party evaluation.
It also mandates the submission of detailed reports on the design, functionality, and intended use of these AI systems to a newly established Office of Artificial Intelligence. The bill also calls for penalties for organizations that failure to comply with the bill’s provisions.
Support for SB-1047
Proponents of SB-1047, including many AI ethics experts and consumer protection advocates, argue that the bill is essential to ensure the safe and responsible development of AI technologies.
The emphasis on pre-deployment testing and independent evaluation helps identify and address potential risks before AI systems are widely deployed. Additionally, the requirement for detailed reporting promotes transparency and accountability, allowing for a better public understanding of AI’s capabilities and limitations.
Opposition to SB-1047
Critics of the bill, including some technology companies, express concerns about the potential for stifling innovation and hindering the development of new AI applications. They argue that the regulatory burden imposed by SB-1047 could disproportionately affect smaller companies and startups, impeding their ability to compete in the AI space.
Additionally, there are concerns that the definition of “frontier AI models” may be too broad, and unclear capturing a wider range of AI systems than intended.
Last, enforcing SB-1047 could prove difficult, especially for non-US based businesses providing AI services to California companies or consumers. Furthermore, as we have discussed, many applications and services are an amalgamation of an ecosystem of services, therefore ensuring all supporting services adhere to these regulations could prove complex.
What Should You Do?
Technology providers developing frontier AI models must start to prioritize robust safety and security measures. Now is the time to invest in comprehensive testing and independent evaluation, ensuring compliance with the bill’s provisions. Transparency and clear communication about the capabilities and limitations of AI systems are crucial for building public trust and demonstrating regulatory compliance efforts.
End-users, including businesses and individuals, must be vigilant about the AI systems they adopt and use. Understanding the potential security, ethical, privacy and accuracy risks and limitations of these systems is essential.
Bottom Line
Regulation of AI applications and services is coming.
Clearly governments and the public have realized that AI technologies have advanced well beyond current regulatory requirements. Technology providers and end-user businesses leveraging AI must get prepared now.
SB-1047, needs further refinement and clarity, and will likely be augmented and enhanced with more regulations over time. The key will be striking the balance between regulation and innovation, which will be an on-going technology, business and regulatory challenge for years. However, SB-1047 does represent a significant step towards regulating the development and deployment of large-scale AI models in California.
Check out Aragon’s Upcoming Webinars
Key Trends Impacting the Evolution of LLMss
It has been over a year since the release of ChatGPT, and now we are seeing an entire market of providers emerge to compete in LLM software offerings. What
do you need to know now?
During this webinar, we will explore emerging trends in large language model development and provide actionable advice for enterprises considering an investment in LLMs.
• Why are LLMs so critical for the evolution of Artificial Intelligence systems?
• What are the major trends impacting the LLM market?
• How will this market continue to evolve over the next 3-5 years?
Join our Experts for Aragon’s September Transform Tour!
Whether you’re a seasoned AI practitioner, a business leader looking to stay ahead of the curve, or simply curious about the future of technology, this virtual event is your chance to gain early access to critical insights that will shape 2025 and beyond.
This isn’t just another trends forecast. Join our expert analysts and women-in-tech panel for:
- Women-In-Tech Guest Panel: Communication in the Age of AI
- Predictions for 2025: The AI Big Bang: From AI Technologies to AI Business Strategies
- AI Assistants: On Your Phone or in the Cloud
Have a Comment on this?