Two days before a Meta scientist gave birth, she received a notification that she had been laid off. Her case is one of twenty-six listed in a 71-page federal complaint filed this week in the Northern District of California, which alleges that the parent company of Facebook, Instagram and WhatsApp used a suite of artificial-intelligence tools to construct its termination list — and that those tools systematically penalised employees who had exercised legal rights to take leave. The lawsuit targets Meta's workforce reduction of approximately 8,000 employees carried out earlier this year. According to the complaint, the company deployed what it calls a "constellation of internal artificial intelligence systems", combining AI-generated performance ratings with keystroke-monitoring and activity-tracking data to score, rank and select workers for dismissal.

The critical flaw alleged by the plaintiffs is structural: when an employee is on approved medical, disability or maternity leave, the very inputs those systems rely on — productivity metrics, activity logs, engagement scores — either do not exist or are artificially depressed. The algorithm, in effect, recorded absence as failure. One engineer stated that he received a lowered performance rating because of time taken off for an injury; a manager on medical leave reported a similar outcome. The plaintiffs argue that Meta's process substituted automated scoring for the "considered judgment of managers who knew the work", converting a legally protected pause into a quantifiable liability.

The case arrives at a moment when regulators across at least three US states — California, Colorado and Illinois — have moved to restrict algorithmic bias in employment decisions, and when similar scrutiny is building in the European Union. Whether a court will halt the layoffs while the claims proceed may itself signal how much institutional weight the law is prepared to place against the machine.