California's new laws set limits on AI workplace monitoring: bathroom breaks, emotions and firings
Employees at Co-op Legal Services in the UK have every single customer call recorded, analyzed and given a percentage score by AI models, according to fresh reporting in The Guardian. That same month, something new happened: on 30 September 2026, California Governor Gavin Newsom signed a package of workplace laws that sets legal boundaries for exactly this kind of monitoring system — most of them taking effect on 1 January 2027. The story shows practice and regulation side by side in the same news cycle.
What the Co-op reporting shows
The Guardian's TechScape newsletter, written by Blake Montgomery and published on 13 October 2026, describes how Co-op Legal Services uses AI models to record, analyze and assign a percentage score to the calls between advisers and customers seeking advice on probate, wills and estates (Guardian TechScape).
The sources in the story are anonymous. An unnamed whistleblower describes the system, according to the Guardian, as "oppressive and dystopian." A Co-op employee who spoke to the paper on condition of anonymity said, according to the reporting: "You are being watched."
Co-op rejects the characterization. The company said, according to the Guardian, that it does not recognize the criticism and that its probate advisers are highly engaged. This is the company's own account, and the contradiction between the workers' anonymous descriptions and the company's version remains unresolved. Neither side's claims have been independently verified.
It is worth noting what the reporting does not contain: neither the vendor, the scoring methodology, nor which employees are covered by the system is specified. What stands firm is the core — recording, automated analysis and a percentage score for every call interaction.
How such systems work — and why they are controversial
The pattern described at Co-op is a familiar category of tool: customer calls are recorded, software analyzes the content, and an automated assessment is produced for every interaction, often as a numeric value that can be compared across employees and over time.
The Guardian itself points to the problem with this model in the same reporting: workers worldwide are increasingly being minutely monitored in the workplace, and small changes in behavior can result in negative performance evaluations. Once the assessment is automated, it is also scalable — a system can assess every call, every day, not just a sample.
The Guardian places the Co-op case within a larger pattern of previously reported complaints, which the paper recounts without re-verifying them: Employees in Amazon's warehouses have previously complained of being timed on bathroom visits. And at Kaiser Permanente, a California hospital chain, nurses have said automated systems assessed their tone of voice in patient interactions. Both examples are therefore second-hand accounts of earlier cases — not newly confirmed documentation.
California responds with legislation
The same month the Guardian reporting was published, regulatory counterweights took shape in the US. On 30 September 2026, Governor Gavin Newsom signed several new employment laws covering artificial intelligence, workplace surveillance and pay reporting, according to a review by the law firm Polsinelli, published on JDSupra on 9 October 2026. Most provisions take effect on 1 January 2027.
Three of the laws touch this monitoring practice directly:
- AB 1331 expands California's workplace privacy protections by restricting employers from using monitoring technology, including certain tracking tools, to monitor employees in workplace bathrooms.
- AB 1883 prohibits AI tools that collect neural data or infer employees' emotional states.
- SB 947, effective 1 July 2027, sets limits on automated decision systems in personnel decisions: employers may not rely solely on an automated decision system to discipline or terminate an employee. If an employer relies primarily on such a system, a human reviewer must corroborate the recommendation using underlying data or other documentation.
SB 947 is the most directly relevant to the practice described at Co-op — even though the law does not apply in the UK. It establishes a principle: a model-generated score cannot alone form the basis for disciplinary action or termination — a human must verify the conclusion against the source data.
What the laws actually cover — and what they don't
It is important to keep the jurisdiction clear. These are state laws in California, with effect for employers there. They do not apply to British companies like Co-op, and they are not national legislation in the US. Co-op's system is therefore currently assessed only under UK regulation, which these sources do not address.
There are no enforcement cases or legal precedents under the new laws yet — most provisions do not take effect until 1 January 2027, and SB 947 not until 1 July 2027. The description of the laws' content comes from a law firm review, not from the statutory text itself.
What remains to be seen
Several central questions remain open:
- Who is right about Co-op? The workers' and whistleblower's descriptions and the company's rejection form an unresolved contradiction. No independent documentation of the system's effect on employees exists in the available material.
- How does the scoring work? Without details on the vendor, methodology and scope, it is impossible to assess how accurate or fair the assessment is — a gap that the California laws only partly address, by requiring human review against the data.
- Will more states follow? California's package is the most concretely formulated regulatory movement in this source material, but its reach beyond the state is unknown.
The Co-op case shows that AI-driven workplace monitoring — from timed bathroom visits to assessed tone of voice in nurses' speech, according to previously reported complaints recounted in the Guardian — also exists in the British service sector, where customers call about the most vulnerable moments of their lives: deaths, estates and wills. California's new laws are concrete legal counterweights — but they apply in one state, and they do not fully take effect until 2027.