"Fired after 17 late arrivals." "Targeted for layoffs after taking maternity leave."
A growing number of U.S. companies are using artificial intelligence to evaluate employees and even single out candidates for dismissal. As AI encroaches on the role of the workplace supervisor, California has moved to bar the technology from firing people without human review.
AI Flags 17 Late Arrivals, Then Recommends Dismissal
In a store-management experiment disclosed on the 5th by U.S. AI startup Andon Labs, an AI manager named Luna judged one employee to be a candidate for dismissal. The worker had arrived late 17 times across 23 shifts.
The system initially issued no warning because it did not recall the relevant rules. Once the company supplied the policy and a record of earlier warnings, it concluded that the worker should be dismissed for repeated tardiness and failure to follow instructions. The final decision and notification were handled by a person.
AI monitoring of employees is already extensive. An AI monitoring program Meta introduced this year was designed to collect keystrokes, mouse movements and browser histories, along with messages, emails and location data from company devices. After more than 1,600 employees signed a petition calling it an invasion of privacy, Chief Executive Mark Zuckerberg halted the program in June.
Such systems are not confined to a handful of companies. In a survey released last year by the Organisation for Economic Co-operation and Development, 90% of U.S. companies were found to use at least one algorithmic management tool to assign work or to monitor and evaluate employees.
Meta Workers Sue Over AI's Role in 8,000 Job Cuts
Claims that AI disadvantaged workers during actual job cuts have also reached the courts. According to The Guardian in July, 26 Meta employees filed suit in the U.S. District Court for the Northern District of California, alleging that the company used AI-based performance reviews and activity monitoring data to select targets for dismissal as it cut about 8,000 jobs.
The complaint cites specific cases. One scientist received a dismissal notice two days before giving birth while on approved pregnancy leave. An engineer who took leave because of an injury said the time away lowered their performance rating, and one manager was fired 16 days after beginning medical leave.
The plaintiffs argued that in scoring productivity and activity levels, the AI effectively treated time not worked because of leave as poor performance. All 26 plaintiffs had taken legally protected leave or had requested or received disability accommodations.
Meta denied the allegations. The company said the claims "are without merit and not based on facts," adding that decisions about workforce management and organization were made by people, not AI.
No Robo Bosses Act Arrives: AI Cannot Fire Workers Alone
As AI moves beyond monitoring and evaluation toward selecting who should be let go, California has applied a legal brake.
According to CNBC and other outlets on the 1st, Governor Gavin Newsom signed the No Robo Bosses Act (SB 947) on Sept. 30. The measure bars companies from firing or disciplining employees based solely on the judgment of an automated decision system (ADS), making it the first legal mechanism in the United States to require human review of AI personnel decisions.
Under the law, even when AI identifies an employee as a candidate for dismissal or discipline, a human manager must review the case again based on work evaluations and performance records. If AI was used as a primary tool in a personnel decision, the company must notify the employee in writing and explain the data used in reaching the judgment. The law takes effect on July 1 next year.
Rules on AI surveillance in the workplace will also be tightened. Restrictions will apply to the use of AI monitoring tools in restrooms and to inferring employees' emotional states from biometric data. The related surveillance law takes effect in January next year, with penalties of up to $500 per violation. As an era arrives in which AI evaluates people and weighs whether to fire them, regulators are moving in earnest to keep the final decision in human hands.
