California signs first US law requiring human review of AI firing decisions
In short
California Governor Gavin Newsom signed the No Robo Bosses Act (SB 947) on September 30, barring companies from firing or disciplining workers based only on automated decision systems and requiring human review. The law follows cases in which AI was used to pick layoff targets, including an Andon Labs experiment in which AI manager “Luna” recommended firing a worker late 17 of 23 shifts, and a lawsuit by 26 Meta employees. Meta denies that people, not AI, made its staffing decisions.
Read the full story 2 min read
California Governor Gavin Newsom signed the No Robo Bosses Act (SB 947) on September 30, three business outlets reported, creating what they described as the first US legal mechanism requiring human review of AI personnel decisions. The law bars companies from firing or disciplining employees based only on the output of an automated decision system and takes effect on July 1 next year. [ 1 , 2 , 3 ]
Under the law, if an automated system's judgment is a main basis for a dismissal or disciplinary action, a human manager must review data, manager evaluations, HR records and the employee's work output to confirm the grounds. If that basis cannot be confirmed, or the system's output is inaccurate, incomplete or misleading, it cannot be used. Companies must notify the employee in writing that an automated system was a main basis and that a person reviewed it, and the employee can request an explanation of the data used. Related monitoring rules take effect in January and restrict the use of AI surveillance tools in restrooms and the inference of employees' emotions from biometric data, with penalties of up to USD 500 per violation. [ 1 , 2 , 3 ]
The signing followed cases of AI involvement in staffing decisions. Andon Labs, a US AI startup, said its AI manager “Luna” recommended firing a worker in an experimental San Francisco store, AndonMarket; the employee had been late 17 of 23 shifts, and Luna also cited violations of work instructions and corporate-card management, according to Asia Economy. Seoul Economic Daily and Maeil Business reported that Luna initially failed to apply company rules and issued no warning until the company supplied the rules and the existing warning record; the final decision and the notification to the employee were handled by a person. Asia Economy dates the announcement to August 14, while the other two outlets date it to October 5. [ 1 , 2 , 3 ]
Separately, 26 Meta employees sued in July in the Northern District of California federal court, alleging that AI-based performance evaluation and activity monitoring data were used to pick targets in about 8,000 layoffs. The plaintiffs cited a scientist notified of dismissal two days before giving birth while on approved maternity leave, an engineer whose rating fell because of time off for an injury, and a manager dismissed 16 days into medical leave. Meta denied the allegations, saying the claims “have no basis and were not grounded in fact” and that decisions about workforce management and organization were made by people, not AI. Meta's AI monitoring program, introduced this year, was designed to collect keyboard input, mouse movement, browser history, messages and emails on company devices and location data; Zuckerberg halted it in June after more than 1,600 employees signed a petition citing privacy violations. [ 1 , 3 ]
The coverage also cites an OECD survey released last year finding that 90% of US companies use at least one algorithmic management tool for work direction, employee monitoring or evaluation. [ 1 , 3 ]
Why it matters
The law is described by the outlets as the first US legal mechanism requiring human review of AI personnel decisions, setting a compliance standard for companies that use algorithmic management tools. The OECD survey cited in the coverage suggests such tools are widespread among US companies, so the rules on written notice and data explanation could affect how employers deploy them. The separate monitoring restrictions, effective earlier, extend the regulation to workplace surveillance.
Key facts
- California Governor Gavin Newsom signed the No Robo Bosses Act (SB 947) on September 30, according to the reports. [ 1 , 2 , 3 ]
- The law bars companies from firing or disciplining employees based solely on an automated decision system and requires human review, with effect from July 1 next year. [ 1 , 2 , 3 ]
- Andon Labs said its AI manager “Luna” recommended firing an employee who was late 17 of 23 shifts, but the company said a human made and delivered the final decision. [ 1 , 2 , 3 ]
- Twenty-six Meta employees sued in July over about 8,000 layoffs, alleging AI-based performance evaluation and monitoring data were used to select those cut; Meta denies the claims. [ 1 , 3 ]
- Meta introduced an AI monitoring program this year and halted it in June after more than 1,600 employees signed a petition against it, the reports said. [ 1 , 3 ]
- The OECD found in a survey last year that 90% of US companies use at least one algorithmic management tool for work direction, monitoring or evaluation. [ 1 , 3 ]
- Monitoring rules effective in January restrict AI surveillance in restrooms and inferring employees' emotions from biometric data, with penalties of up to USD 500 per violation. [ 1 , 3 ]
Confirmed by several sources
- California Governor Gavin Newsom signed the No Robo Bosses Act (SB 947) on September 30. [ 1 , 2 , 3 ]
- The law prohibits firing or disciplining employees based only on an automated decision system and requires human review; it takes effect on July 1 next year. [ 1 , 2 , 3 ]
- An Andon Labs AI manager named “Luna” judged one employee a firing target after 17 late arrivals in 23 shifts, with a human making the final call. [ 1 , 2 , 3 ]
- Meta says its workforce and organizational decisions were made by people, not AI, and rejects the lawsuit's claims. [ 1 , 2 , 3 ]
Still unclear
- When Andon Labs disclosed the “Luna” experiment. Seoul Economic Daily and Maeil Business say the disclosure came on October 5, while Asia Economy says Andon Labs announced the recommendation on August 14.
- Whether the Andon Labs store experiment involved real employment decisions. The documents describe an experimental store and an AI manager's recommendation, but do not say whether the employee was actually dismissed.
- Where the law was signed. The documents name California and Governor Gavin Newsom but give no city.
What local media are saying
Timeline, local time
- September 30: California Governor Gavin Newsom signs the No Robo Bosses Act (SB 947). [ 1 , 2 , 3 ]
- August 14: Andon Labs announces that its AI manager “Luna” recommended firing a store employee, according to Asia Economy. [ 2 ]
- October 5: Andon Labs discloses details of the AI-manager store experiment, according to Seoul Economic Daily and Maeil Business. [ 1 , 3 ]
- June: Meta halts its AI monitoring program after more than 1,600 employees sign a petition against it. [ 1 , 3 ]
- July: Twenty-six Meta employees file suit in the Northern District of California federal court over AI use in about 8,000 layoffs. [ 1 , 3 ]
- January (next year): The related workplace AI monitoring rules take effect, with penalties of up to USD 500 per violation. [ 1 , 3 ]
- July 1 (next year): The No Robo Bosses Act takes effect. [ 1 , 2 , 3 ]