California's No Robo Bosses Act, Lyft Settlement & AI Hiring Bias

California has introduced new rules: AI can assist with employment decisions, but who is ultimately responsible? What warning does Lyft's record $272.5 million settlement send to businesses that classify gig workers as independent contractors? And why can AI hiring systems, originally expected to reduce human bias, develop new biases of their own? This report delivers a comprehensive evaluation.

California's No Robo Bosses Act: Can AI Fire Employees?

First, let's look at California's new legislation. On September 30, 2026, Governor Gavin Newsom signed SB 947, the revised No Robo Bosses Act, which will take effect on July 1, 2027. (opens in new tab)

The law primarily addresses employers' use of automated decision-making systems in employee disciplinary actions and terminations. It prohibits employers from relying entirely on automated systems to make these decisions. When an employer primarily relies on an automated system's assessment, two key requirements apply:

1. Independent human review is required. Employers must verify the system's findings using additional information, such as management evaluations, personnel records, employee work performance, coworker feedback, or witness interviews. If the system's conclusions cannot be verified, or if the reviewer identifies inaccurate, incomplete, or misleading information, the employer cannot use those findings as the basis for disciplinary action or termination.

2. Written notice must be provided.Affected employees must receive written notification explaining that an automated system was used and that a human review was conducted. The notice must also include relevant contact information and inform employees of their right to obtain details about the personal data used by the system.

Under the new law, systems that use artificial intelligence, machine learning, or data analytics to generate scores, classifications, or recommendations that assist or replace human judgment may fall within its scope. However, conventional tools such as basic calculators, databases, and firewalls are excluded from the definition.

For employers, human review cannot be a mere rubber stamp.If automated tools are used to handle related matters, employers need to establish corresponding procedures, independently verify the underlying information, retain review records, and send written notices to employees as required.

California Announces $272.5 Million Lyft Wage Settlement

Next, let's look at a record-setting labor settlement. On October 1, 2026, the California Labor Commissioner's Office announced a $272.5 million wage settlement with rideshare company Lyft (opens in new tab)(plus interest), setting a new record for the largest settlement amount in a California wage and hour case.

The lawsuit began in August 2020, with regulators alleging that Lyft misclassified drivers who should have been deemed employees as independent contractors, depriving them of statutory protections such as minimum wage, overtime pay, expense reimbursement, and paid sick leave. The settlement covers the period from April 6, 2016, to December 15, 2020. Note that California's Proposition 22, which governs rideshare driver classification, took effect after the period covered by this case, so the settlement does not require Lyft to reclassify subsequent drivers or cover compensation for drivers after that period.

Regarding the payout arrangements (opens in new tab), 87% of the settlement funds will be distributed directly to drivers. The Labor Commissioner's Office will transfer its entire $5.45 million penalty share to drivers who filed wage claims; more than 1,600 drivers who filed claims will also receive additional compensation, with their payout mileage calculation base doubled. The settlement funds are held in a third-party escrow, and eligible drivers will receive official notices and can get assistance through a dedicated website, email, and call center.

This case reminds business managers that employee classification can involve wage, hour, and other labor protection issues, and related disputes may carry significant financial risk.

AI Hiring Systems Can Develop New Biases

Beyond employment risks, a noteworthy new finding has emerged regarding AI recruiting. A joint study by Princeton University and the University of Chicago found that large language models may develop new social biases on their own when repeatedly making hiring decisions (opens in new tab).

The experiment set up four groups of fictional job candidates with identical success probabilities. After repeated simulated hiring rounds, the large model spontaneously developed differentiated career tendencies for different groups. The researchers explained this mechanism using the "exploration-exploitation trade-off": once the AI happened to receive positive feedback, it would keep reusing that decision pattern, gradually hardening a single chance judgment into a new stereotype.

However, the experiment was a controlled simulation using fictional candidates, with the model receiving immediate feedback on hiring outcomes; real-world recruiting tools mostly cannot track candidates' actual performance after hiring. Therefore, the extent to which this bias affects real-world recruiting still requires further research.

The study also found that approaches such as adjusting prompt randomness or shortening historical reference records had limited effect in suppressing new biases. Compared with vaguely instructing AI to remain fair, directly embedding diversity goals in the instructions was more effective at reducing stereotype formation.

This research shows that AI tools cannot simply be left to run on their own once deployed. In addition to fairness assessments before deployment, companies need to continuously monitor system performance, regularly review outputs, conduct human review when necessary, and promptly identify any new biases that may emerge.

KCAL Insurance Agency's Manager Weekly: five minutes every week bringing you the latest updates on US business operations and talent management!

This article is for general informational purposes only and should not be considered legal, tax, or accounting advice. We recommend seeking specific professional advice and assistance.Click here to schedule a free business legal consultation. (opens in new tab)