By 
A lawsuit filed in a California federal court in July has become the clearest test yet of what happens when companies let software make consequential decisions about people's jobs. Twenty six current and former Meta employees are suing the company in the U.S. District Court for the Northern District of California, alleging that an internal AI monitoring system flagged them for layoffs largely because they were on approved medical, parental or pregnancy leave. The case lands at the same moment Kenya's Senate is reviewing its own Artificial Intelligence Bill 2026, legislation that would, on paper, give Kenyan workers some of the protections the Meta plaintiffs are now trying to establish through litigation.
The lawsuit, Does 1-26 v. Meta Platforms Inc, was filed as part of the roughly 8,000 person layoff Meta carried out starting in May. According to the complaint, Meta tracked employee activity through keystroke data, mouse movement, AI token usage dashboards, browser activity and algorithmically generated performance rankings. The plaintiffs argue that these metrics, by design, could not be accumulated by anyone on protected leave or whose output was reduced by a disability, and that Meta did not adjust the scoring system to account for that before using it to help decide who would be let go.
About half of the 26 plaintiffs had taken leave connected to pregnancy, caregiving or a medical condition. Eight are women who had taken maternity or pregnancy related leave, four are men who had taken parental leave, and one plaintiff took leave to care for a family member and later for bereavement. One research scientist named in reporting on the case was notified of her layoff two days before giving birth, while she was on pre birth pregnancy leave. The suit invokes the Family and Medical Leave Act, the Americans with Disabilities Act, the Pregnancy Discrimination Act and the Pregnant Workers Fairness Act, along with the disparate impact doctrine established in a 1971 Supreme Court ruling, arguing that a facially neutral scoring system produced a discriminatory outcome because women disproportionately take pregnancy and caregiving leave. Meta has said the claims lack merit and maintains that human managers, not AI, made the final layoff decisions. The company has not responded to further requests for comment beyond that statement.
Not an Isolated Case
The Meta suit does not stand alone. Workday, whose software is used by thousands of employers to manage hiring and workforce decisions, is defending a separate case, Mobley v. Workday, alleging that its AI powered applicant screening tools discriminated against candidates on the basis of race, age and disability. Workday has denied the allegations and says hiring decisions are made by its customers, not by its algorithms. A Bloomberg investigation earlier this year also found indications that OpenAI's ChatGPT produced different outcomes for identical resumes depending on the name attached to the applicant, raising questions about bias embedded in general purpose AI tools that companies are beginning to fold into recruitment.
The tension is not limited to companies filtering candidates out. Google's Workspace division markets its AI features to business customers as a way to draft job postings, evaluate resumes and forecast hiring needs, positioning automation as a time saving tool for HR teams. At the same time, Google DeepMind's AGI Safety and Alignment team has taken the opposite approach internally. According to a document reported by Bloomberg, the team created a supplementary application form specifically to help candidates bypass Google's own automated screening systems, warning that there is a non trivial chance a CV will be incorrectly filtered out or delayed. The form tells applicants plainly that a real human will read their answers and that reviewers grow tired of responses that read as though they were generated by a language model, because AI generated answers tend to sound alike. The disclaimer on the internal document read "please do not share this doc widely," an acknowledgment that the company's own AI recruiting tools were seen internally as unreliable enough to need a workaround.
Where Kenya's Bill Fits
Kenya does not yet have a Meta scale case in its courts, but the underlying pattern, automated systems making or shaping decisions that affect people's livelihoods without a clear route to explanation or human review, is exactly what the Artificial Intelligence Bill 2026 attempts to regulate. Sponsored by nominated Senator Karen Nyamu and currently before the Senate, the bill would create a Right to Explanation for anyone affected by a significant automated decision, including a job application screening or a loan rejection, along with a Right to Human Review and a Right to Challenge before a decision is finalised. Had an equivalent framework existed in the United States, it would have given the Meta plaintiffs a statutory route to demand a human review of their scores before termination, rather than relying on post hoc litigation under employment discrimination law.
The bill is not without problems, a point TechInKenya has previously examined in detail. Its requirement that developers produce audit trails showing how an underlying AI model was trained is workable for a company that builds a foundation model from scratch, but most Kenyan AI developers do not do that. They fine tune existing open source models such as Llama, Mistral or Gemma for local use cases like Swahili chatbots, mobile money fraud detection or agricultural diagnosis tools. The training data decisions behind those base models were made by researchers at Meta or Google, not by the Kenyan developer adapting them, which means the audit trail the bill demands does not exist in a form local developers can produce. Critics, including analysts cited by Business Daily's editorial board, have also flagged the bill's three new institutions, the Office of the AI Commissioner, the AI Authority and the AI Advisory Council, as a potential source of duplication given that the Office of the Data Protection Commissioner and the Communications Authority already have overlapping mandates.
Those drafting issues matter because they determine whether the rights on paper translate into rights that can actually be enforced. A Right to Explanation is only meaningful if the entity deploying an AI hiring or monitoring tool can be compelled to produce one, and if the penalty structure does not simply push smaller developers out of compliance altogether. The Meta and Workday cases give Kenyan lawmakers a live, well documented example of what happens when neither of those conditions is met: workers left arguing after the fact that a system they never had visibility into made a decision no human meaningfully reviewed. The Senate's ICT Committee, which has been taking public input on the bill, will need to decide whether the audit trail and institutional design questions get resolved before the bill moves further, or whether Kenya ends up with a framework that reads well but proves as difficult to enforce as the current gap it is meant to close.
Comments