Shaping AI’s effect on employment in California

We are in a short lull in the legislative battle between Big Tech and Big Labor over the impact of AI on jobs in California. But when lawmakers return in a few weeks, they will deal with a plethora of measures aimed at how artificial intelligence is utilized the workplace.

First up are a number of bills sponsored by the California Labor Federation of Labor Unions. The proposals range from requiring humans to be involved in discipline, termination and deactivation decisions, allowing healthcare workers to rely on their professional judgement, and requiring 60 days’ notice before technological displacement to affected employees and government departments.

“These efforts reflect a broader principle: California should continue leading in innovation while also leading in responsible governance,” Senator Eloise Gómez Reyes’s (D-San Bernardino) office said in a statement to Capitol Weekly. “As AI evolves, policymakers need reliable information and thoughtful safeguards so technological advancement benefits Californians without leaving workers or communities behind.”

Many of the bills are opposed by the California Chamber of Commerce and a coalition of corporate interests concerned that the proposed regulations would have a “chilling effect” on the development of AI and the benefits they say the technology brings to the California economy.

Here are a few of the measures that have been moving through the Legislature.

SB 947
Authored by Senator Jerry McNerney (D-Pleasanton) and co-authored by Reyes and Assemblymember Chris Ward (D-San Diego), SB 947, the No Robo Bosses Act, would limit the way employers in California can use automated decision systems, or ADS, to make decisions on disciplining, terminating or deactivating employees and workers. In this case, deactivation refers to the accounts of app-based workers, such as rideshare and delivery drivers.

“It concerns me that we would be using these automated decision-making systems to discipline and fire people without a human being in the loop,” McNerney said. “They can be used, and they can be helpful, but I want to make sure there’s human beings in the loop because we’re using these systems that have a great impact on people’s lives.”

McNerney said that it is important to keep humans in the loop because they have “understanding and compassion,” which bring perspective to an employee’s performance at work.

“They’re not going to use just hard and cold data,” McNerney said. “Human beings can understand how a worker is being productive even if they miss certain metrics.”

While the Chamber acknowledges concerns about “employees being unfairly disciplined or terminated” by automated tools, it believes the bill “does not consider the benefits of ADS technology.”

“The bill broadly targets businesses of all sizes, across every industry, and regulates even low-risk applications of automated decision systems,” the Chamber wrote in a letter of opposition to the measure. “SB 947 will drive up costs for consumers and employers because it would impose significant compliance burdens and any misstep would lead to costly litigation for even the smallest of employers.”

“It concerns me that we would be using these automated decision-making systems to discipline and fire people without a human being in the loop.”

The bill is also opposed by the California Special District Association, which is concerned that the bill’s mandate will divert public-sector resources at a time of budget constraints without a clear need.

“We’ve asked all the sponsors to provide us with concrete, specific examples of public employers abusing these types of technologies, and we’ve been provided with none to date,” said Aaron Avery, CSDA’s Director of State Legislative Affairs. “So we’re left with kind of a private sector issue and a hypothetical problem in the public sector workforce that, unfortunately, the solution would expose us to litigation.”

The Chamber said SB 947 would also undo many of the compromises made in last year’s SB 7, a version of the bill that was ultimately vetoed by Gov. Gavin Newsom, “marking a step backwards in progress on this issue.”

McNerney said his office has made many compromises on the current bill, “taking into consideration what the biggest concerns are of employers.”

“We don’t want this to be something that costs them a tremendous amount of money or lawsuits,” McNerney said. “We want these systems to be used; we just want them to be used with human oversight.”

Lorena Gonzalez, president of the California Federation of Labor Unions (shorthanded by most Capitol community denizens to Cal Labor Fed), said the politics of how people view AI have changed since the veto, with the governor more likely to take a second look at the bill.

“You can never know what the governor is going to do,” Gonzalez said. “We’re working closely with the administration, as we did last year, to try to ensure that we get a bill that they can recommend the governor to sign.”

The Chamber also opposes SB 947’s requirement that a human review and corroborate violations that lead to discipline, termination, or deactivation decisions, stating that such corroboration “may be impossible or highly impractical.”

Avery said that CSDA is “open to humans being required to make final decisions,” but by and large, humans are already in the loop.

The Chamber also rejects SB 947’s other employee protections, saying the bill is “at odds with prior legislation and court decisions.” The Chamber argues that ADS is used “to flag inappropriate behavior, potential fraud, or other issues” that would lead to immediately terminating an independent contractor, something that would be prohibited under SB 947, “even if that termination is necessary for employee or consumer safety, to stop a seller from selling counterfeit goods or defrauding customers, and more.”

“We want to make sure that contract employees are treated fairly as well,” McNerney said on their inclusion. “Those are the employees that probably need the most amount of protection.”

SB 947 would also require employers who used ADS to make a disciplinary, termination, or deactivation decision to issue a post-use notice informing the employee that ADS was primarily used in the decision and that it was reviewed by a human. Additionally, employees would be able to request a copy of the most recent 12 months of data used in the ADS decision.

But the Chamber also raised issue with those requirements, saying they risk undermining the integrity of investigations that rely on confidential communications with “other workers, HR or legal counsel involvement.”

Sarah Dukett, Policy Advocate at Rural County Representatives of California, said such requirements also place a financial burden on local governments.

“This would be an example of a mandate that we would be required to fulfill, but we would get no state funding,” Dukett said. “We have to figure out how to implement these provisions within the existing resources we have.”

SB 947 also bans using ADS to predict employee behavior for employment-related decisions, with McNerney saying “people shouldn’t be punished for something they haven’t done.”

But the Chamber claims the ban would be too broad, stating that SB 947 would hinder financial institutions that “sometimes use ADS for predictive purposes for assessing risk of fraud or other unlawful activities.”

Avery said the ban would impose restrictions on the public sector for what is a private-sector issue.

“This would be an example of a mandate that we would be required to fulfill, but we would get no state funding,” Dukett said. “We have to figure out how to implement these provisions within the existing resources we have.”

“We would need to be provided with an example first of our public employers using that type of predictive technology in a way that is concerned with in order to really directly address that,” Avery said. “We’ve not been provided with examples.”

SB 947 passed the Assembly Judiciary Committee on an 8-3 vote and will be heard next in the Assembly Appropriations Committee in August.

SB 951
Authored by Reyes, SB 951, known as the California Worker Technological Displacement Act, requires employers to provide a 60-day notice to affected workers, the Employment Development Department and local entities before a worker can be displaced by the adoption of AI and other automating technology. Local entities include workforce investment boards, city councils and county boards of supervisors of each city and county with affected workers.

Under SB 951, employers with 100 or more employees would be required to give workers affected by technological displacement the right to first bid on an open position with the employer.

In an email to Capitol Weekly, Reyes said the question is no longer whether AI will change the economy, but whether the transition will be managed responsibly.

“SB 951 is about transparency, accountability, and giving workers and policymakers the information they need to prepare for the future….For employees, that can mean additional time to seek new employment, explore retraining opportunities, or access workforce services,” Reyes said in the email. “For the Employment Development Department, it provides earlier awareness of significant workforce changes so existing resources and programs can better respond to affected communities.”

The Chamber said the bill is premature following Gov. Newsom issuing Executive Order N-6-26 that called for a review and update of California’s worker protections, such as the WARN Act, in response to AI and automation.

“The Governor directed LWDA to review the WARN Act, assess existing worker protection programs, and develop workforce data tools, all through a structured process involving labor, employers, and academic experts due later this year,” the Chamber said in an opposition letter sent in June to the Assembly Labor and Employment Committee. “Passing SB 951 now would bypass the Executive Order’s structured review process, the very process designed to determine whether any legislative change is needed in the first place.”

Dukett, who leads a coalition of local government interests against SB 951, said the bill is “another example of public sectors being put into a private sector process.”

“This bill tries to create a new type of noticing requirement under CalWARN,” Dukett said. “Public agencies are not required to comply with CalWARN today; we do not do noticing under those provisions.”

The Cal Labor Fed’s Gonzalez believes data is needed from both the private and public sectors.

“We need accurate data, whether it’s public sector or private sector, on which jobs are being lost to AI,” she said. “Those are things that we have to know; we have to know to be able to plan for our future.”

Dukett said that the bill would place a private-sector process over the public sector, even though the two are generally regulated differently. She added that some of the bill’s provisions don’t fit with the heavily unionized nature of the public sector and the extensive notifications before public-sector layoffs.

“We’re also concerned that the bill treats independent contractors as employees,” Dukett said. “Independent contractors are not employees under the government code, and this would be something that would be completely new, and in many ways we think that’s very inappropriate in terms of how we contract with these people; we have contracts in all of our contracts so that at any time we can end them.”

The Chamber also cited concerns over SB 951 including independent contractors with six months of tenure in the bill’s protections and “overlapping and potentially inconsistent obligations” between SB 951 and the WARN Act when it comes to the right of first bid.

“They’ve spent hundreds of millions of dollars to try to write their own laws because they want no laws to basically adhere to them.”

Gonzales, who during her time in the Assembly authored AB 5, which sets a standard for determining whether a worker is an independent contractor or an employee, said this isn’t about employers with one or two contractors, but about employers with massive numbers of contractors whom they treat as employees.

“They’ve spent hundreds of millions of dollars to try to write their own laws because they want no laws to basically adhere to them,” Gonzalez said. “These are companies that have decided to use a business model to avoid the liabilities and labor laws, but they shouldn’t be able to get out of some basic provisions like these.”

SB 951 was endorsed in the Assembly Committee on Privacy and Consumer Protections and is now in the Appropriations Committee.

AB 1883
Authored by Assemblymember Isaac Bryan (D-Los Angeles), AB 1883 would ban AI surveillance tools in the workplace that “collect neural data or recognize an individual’s emotional state.”

Not surprisingly, this has also drawn the ire of the Cal Labor Fed’s Gonzalez.

“We call it an AI strip-search,” Gonzalez said. “We want to make sure that people’s neural brain waves aren’t being taken by their employer and being analyzed.”

Eric Lawyer, a senior legislative advocate at the California State Association of Counties, said Bryan and the California Federation of Labor Unions have been responsive to CSAC’s and a coalition of public employers’ concerns.

“From the beginning with AB 1883, our chief concern has been a possibility that it could restrict the use of essential security system tools,” Lawyer said. “We opposed the earlier version of the bill because of a possibility that it could restrict pretty essential security tools, specifically that use facial recognition.”

Gonzalez said opponents of the bill raise the notion that modern security cameras are AI-enabled, but that it shouldn’t be used as a red herring.

“We want to make sure that people can still feel safe and that that’s not used as a red herring,” Gonzalez said. “Cameras that have existed in workplaces for a very long time can still be used to ensure that they’re safe without being used as an additional tool of mass surveillance.”

During AB 1883’s hearing before the Senate Committee on Labor, Public Employment and Retirement, Andrea Lynch, Policy Advocate for the Chamber, said the bill conflicted with SB 533’s requirement “to have effective means to alert employees to a threat in real time.”

“Many use AI-enabled security systems that detect raised voices, distress, escalating aggression, or facial recognition in workplaces or parking lots,” Lynch said. “This is to be able to trigger an alert before a situation turns violent. This is precisely the kind of emotional inference AB 1883 prohibits.”

Lawyer said they are currently reevaluating the bill over the summer recess after amendments were introduced that narrow its focus to tools for monitoring neural data and emotional states.

Lawyer added that a broader concern with AB 1883 and other workplace technology bills is the inclusion of the public sector, which is already required to work with its workers’ unions on employment matters.

“Many use AI-enabled security systems that detect raised voices, distress, escalating aggression, or facial recognition in workplaces or parking lots,” Lynch said. “This is to be able to trigger an alert before a situation turns violent. This is precisely the kind of emotional inference AB 1883 prohibits.”

“I think at its core our concern has been any restrictions on tools that we use to keep our public sector workforce, our public officials and the public we serve safe,” Lawyer said. “We have concerns about any legislation that overly restricts or limits the use of new technologies that we use to ensure that agencies can better provide essential services.”

AB 1883 passed the Senate Committee on Labor, Public Employment and Retirement and sent to the Senate Committee on Appropriations.

AB 2575
Authored by Assemblymember Liz Ortega (D-Hayward), AB 2575 would protect health care workers who use their professional judgement in the direct treatment of patients over AI recommendations.

“AB 2575 protects healthcare workers from retaliation when they speak up when it comes to AI and the tools they’re being asked to use,” Ortega said. “They recognize that sometimes they get it wrong, and they don’t want to be reprimanded or retaliated against when they use their own clinical judgement.”

George Soares, a legislative advocate with the California Medical Association, said that supporters of the bill are operating under a false presumption that healthcare workers are not allowed to use their professional judgement.

“We have not seen that or heard that from our members at all, and we’ve double- and triple-checked this year as well,” Soares said. “I think that that is something that is fundamentally flawed in their overall direction. There is not this rampant direction from physicians or hospitals directing health care workers to just go with whatever the AI tool says.”

Gonzalez said it’s a matter of the education health care workers have received versus the unknown of AI.

“The professional judgement of a nurse has been trained and tested where the computers that are being brought in, we don’t know that they’ve actually been trained and tested,” Gonzalez said. “Professional judgement is much easier to attribute to human beings who have actually gone through the training necessary by the state of California to practice health care.”

Soares said the bill could be used to give nurses the power to override decisions in cases in which it would not be appropriate.

“It undermines the overall delivery of care to the patient, which is typically governed by the attending physician,” Soares said.

According to Soares, CMA agrees with keeping a human in the loop and ensuring AI is tested for better use, but said that other bills, like AB 1979 and SB 503, better capture the intent of AB 2575.

“I think that would show this bill is somewhat unnecessary,” Soares said.

Ortega said her bill complements those bills by offering protections to healthcare workers.

“I’ve worked on this bill pretty substantially this year, taking into consideration a lot of the feedback that I received from experts in the field and my colleagues,” Ortega said. “Unfortunately, the California Medical Association has never sent us any suggestions or feedback on the bill.”

In its opposition letter, the Chamber said the bill would “ultimately reduce patient access to beneficial technology.”

“AB 2575 fails to distinguish between AI that informs clinical decisions and AI that replaces them,” the letter said. “A licensed professional can, and should, use their professional judgement when using AI tools.”

Ortega said her bill is not about removing AI from healthcare, but about providing accountability when it is used.

“My dad had a massive heart attack last year, and I got to see firsthand the benefits of AI,” Ortega said on her bill. “At the same time, I also saw a lot of things that weren’t explained. I had a lot of questions, I saw a black box that I think many voters see.”

AB 2575 is set to be heard by the Senate Committee on Appropriations on August 3rd.

SB 928
Authored by Senator Sabrina Cervantes (D-Riverside), SB 928 bans the California State University from replacing faculty with AI, while still allowing for its use. Cervantes said AI tools should support educators, not replace them.

“SB 928 is needed to protect the human connection that our students need to learn well,” Cervantes said in an email to Capitol Weekly. “My bill ensures that CSU students will continue to be taught by human instructors.”

SB 928 has received no opposition as it has moved through the legislature.

“I think people are embarrassed to suggest that teachers, faculty and people who inform us should be replaced by AI,” Gonzalez said. “Even the AI companies are embarrassed to have that position.”

Gonzalez said the California Federation of Labor Unions’ goal is to identify professions where it would be inappropriate to replace humans with AI and to introduce bills to protect human knowledge and interaction in those professions.

SB 928 passed the Assembly Higher Education Committee on a 10-0 vote and is awaiting a final vote on the Assembly floor.

Cory Jaynes is a Sacramento-based freelance reporter and a regular contributor to Capitol Weekly.