A Glance at Proposed AI Bills in California

California wants to rein in artificial intelligence—and that means employers and businesses need to step up and pay attention. State lawmakers are considering more than 30 AI-related bills across a wide range of topics while the state’s privacy agency is considering ambitious regulations targeting AI’s impact on California consumers. California’s landmark proposals could inspire similar action in other states, so businesses across the country will want to take note. This article outlines the proposed bills and regulations that employers should care about most.  

5 Bills You Should Know About

Prohibiting Algorithmic Discrimination—AB 2930 (Bauer-Kahan)

If this bill is enacted, employers would be prohibited from using automated decision tools to make a “consequential decision” that results in “algorithmic discrimination.” As a result, a prosecutor could bring a civil action against an employer and seek a $25,000 fine per violation if: 1) the employer uses an automated system to make decisions regarding pay, promotions, hiring, termination or automated task collection; and 2) the automated system contributes to unjustified treatment or impacts disfavoring employees based on race, color, ethnicity or other legally protected classification.

Employers also would be required to notify individuals in advance that automated systems will be used to make consequential decisions. If “technically feasible” (a phrase which has yet to be defined by the bill), employers would be required to accommodate an individual’s request to be subject to an alternative selection process, such as human decision-making.

This bill is a reintroduction of a 2023 bill that did not make it out of the legislature. This time around, the author has some high-profile industry support for the bill, so it could fare better this year. 

Addressing Employment Replaced by AI—AB 3058 (Low)

This bill states that lawmakers intend to enact legislation to create an “unconditional” benefit pilot program for California residents whose employment is replaced by artificial intelligence.

Requiring Safety Standards for Large AI Models—SB 1047 (Wiener)

As perhaps the broadest measure on the table, you’ll be hearing a lot about this one. The “Safe and Secure Innovation for Frontier Artificial Intelligence Models Act” would, among other things, require developers of “covered models” (which the bill defines broadly but targets the handful of extremely large and most powerful AI developers) to:

• Make “positive safety determinations” before releasing covered models to the public (which essentially requires ruling out any possibility of a model having a “hazardous capability,” such as the power to create biological or nuclear weapons or the ability to enable $500 million of damage through cyberattacks).
• Implement protections to prevent unauthorized access to the models.
• Design the models to be capable of promptly enacting a full shutdown until a positive safety determination is reached.  

The bill also protects whistleblower employees who report noncompliance with the act.

Even if you’re not a tech developer, the act’s potential to impact the availability and functionality of AI tools could affect any employer or employee who uses these tools at work. From a broader standpoint, the act might signal how the state views this technology and the direction lawmakers are headed in addressing it.

Tackling ‘Digital Replicas’ in the Entertainment Industry—AB 2602 (Kalra)

The entertainment industry regularly creates and uses digital replicas—one of the hot issues during the actor’s strike last year—of actors in a variety of ways. This bill would make a digital replica contract provision retroactively unenforceable if it:

1) permits creating and using a digital replica of an individual’s voice or likeness either in place of the work of the individual; and 2) does not clearly spell out the proposed uses of the digital replica.

The bill targets situations where there is an imbalance of power, such as when the individual who is at risk of losing work because of a digital replica is not represented by legal counsel or a labor union. The party with the power to create or use digital replicas would be required to notify the individual by Feb. 1, 2025, that a prohibited provision is unenforceable.

Regulating the Use of AI by State Agencies—SB 896 (Dodd)

The “Artificial Intelligence Accountability Act” focuses on the use of AI by state agencies. Among other things, the bill would require various agencies to produce reports on the state’s potential best uses of generative AI tools and perform a joint risk analysis of AI’s potential threats to California’s critical energy infrastructure. The act also would require state agencies to notify people when generative AI is being used to communicate their interaction.

What Other Bills are Floating Around?

Here’s a sampling of other types of bills that California lawmakers are considering.

• Crackdown on Deepfakes. Under proposed legislation, any AI-generating entity or individual that creates a deepfake using a person’s name, voice, signature, photograph or likeness, without that person’s permission, would be liable for any actual damages suffered by that person. Another bill would require the establishment of a working group to research and report on the relevant issues and impacts of AI and deepfakes.
• Election Integrity. The legislature introduced several bills specifically targeting deepfakes used to influence an election or solicit funds for a candidate or campaign. If passed, any materially deceptive and digitally altered or digitally created images or audio/video files would be banned within 120 days of an election (and, in some situations, 60 days after an election).
• AI Usage for Lawyers. Another bill would express the legislature’s intent to enact laws requiring legal professionals to disclose to the court whether they used AI or machine learning to prepare court documents in California and ensure that citations used within those documents conform to ethical and professional standards.
• Preserving Physician-Patient Relationships. If enacted, the “Physicians Make Decisions Act” would require physician oversight when algorithms are used to determine if or what kind of healthcare a patient receives.

What’s Next?

The 2024 legislative session will run through Aug. 31, although we’ll have a much better view of the chances of any of these proposals becoming law by later this summer.

What’s Happening on the Regulatory Front?

The California Privacy Protection Agency’s proposed regulations would broadly protect consumers—Including employees and job applicants—from businesses’ use of automated decision-making technology (ADMT). If finalized, businesses would be required to:

• Notify consumers in advance about the use of the technology, the decision-making process, and the consumer’s opt-out (and access-to-information) rights;
• Allow consumers to opt out of the use of ADMT for decisions that significantly impact their lives, such as decisions about an employee’s compensation, allocation or assignment of work, or hiring decisions; and
• Provide consumers access to information about the use of ADMT.

Formal rulemaking is underway. Many industry groups have expressed concerns that the proposed regulations extend beyond the agency’s authority to regulate privacy issues. The California Privacy Protection Agency’s board is likely to continue revising the scope of these rules during the formal rulemaking process.

Benjamin M. Ebbink is an attorney with Fisher Phillips in Sacramento, Calif., and Washington, D.C. Usama Kahf is an attorney with Fisher Phillips in Irvine, Calif. Kile E. Marks in an attorney with Fisher Phillips in San Diego. Annie Ziesing is an attorney with Fisher Phillips in New York City. © 2024 Fisher Phillips. All rights reserved. Reposted with permission.