180 incidents on record · 2026 Headlights Incident reports by Ellie Harris · Melbourne
10 new this week Library last updated 30 August 2026
← The incident library
HD-INC-155
Logistics · United States · 2024 · Automated decision without human review

California's Labour Commissioner cited Amazon nearly six million dollars for failing to meet a warehouse quota law written three years earlier, at two of its own fulfilment centres

By Ellie Harris · Filed Citations issued June 2024; the quota law took effect 1 January 2022

Alleged: Amazon.com, Inc. developed or deployed the AI system implicated in this incident. Details are drawn from public reports; parties are presumed innocent of any wrongdoing not established by an official finding.

California's Labour Commissioner cited Amazon nearly six million dollars for failing to meet a warehouse quota law written three years earlier, at two of its own fulfilment centres

What happened

It was reported that on 18 June 2024 California’s Labour Commissioner announced citations against Amazon totalling nearly six million dollars. The Commissioner found that Amazon had failed to meet requirements of the Warehouse Quotas law at warehouses in Riverside and San Bernardino Counties. Assembly Bill 701 was enacted in 2021 and took effect on 1 January 2022. It requires warehouse employers to give workers written notice of the quotas they must meet, including the required number of tasks per hour and any discipline for failing to meet them. Under the law a quota cannot require a worker to miss meal or rest periods, give up reasonable access to bathroom facilities, or fall short of occupational health and safety standards. Contemporary reporting widely identified Amazon as the legislation’s principal target.

It was reported that Amazon’s warehouse productivity process had been described years earlier. In April 2019 The Verge reported on a letter dated 4 September 2018 and related documents that it said it obtained through a Freedom of Information Act request. According to that reporting an Amazon lawyer told the National Labor Relations Board that approximately 300 full-time employees at the Baltimore fulfilment centre were terminated for failing to meet productivity or quality targets between August 2017 and September 2018, about ten per cent of the site’s workforce. The documents reportedly described a system that generated productivity and quality warnings or terminations without supervisor input, using a measure called time off task for periods when a worker was not scanning. Amazon rejected that account and said managers can intervene. The underlying letter has not been published and must be requested from the National Labor Relations Board, so the account rests on reporting by The Verge and CBS News rather than the document itself.

What an auditable version would have shown

The California finding concerned whether workers received written notice of what they were expected to do and what would follow if they did not. An auditable version would answer that from the organisation’s records rather than an inspection. For each worker it would retain the quota in force, when and how it was disclosed, every warning generated by the productivity count, and the identity and reasoning of the person responsible for any consequential decision. Amazon says a manager can intervene. A record of this kind would make that claim checkable by showing, case by case, whether a manager did intervene when review was required.

Where the gap was

It was reported that a productivity count generated warnings and terminations, on the account of the documents The Verge obtained. Years later the Labour Commissioner found that workers at two sites had not received the written notices required by law. A ConductRecord preserves what the system did and how any person responded, with enough detail to show whether review was meaningful and whether a manager intervened when needed. A MetricRecord converts individual events into continuing figures, allowing the organisation to monitor how often the count produces warnings and terminations and which workers are affected. The practice had been reported since 2019 and governed by legislation since 2021, rather than first emerging with the 2024 citations.

What governance should have looked like

It was reported that California enacted the written-notice requirement in 2021, with effect from January 2022. In June 2024 the Labour Commissioner found that Amazon had not met it at two warehouses. When an automated measure of work can lead to discipline, the employer should disclose the standard in writing and retain a record showing what a person decided in every consequential case. Amazon disputes the account of automatic termination and says managers can intervene. The sources reviewed identified no regulatory finding outside California concerning Amazon’s algorithmic management of warehouse workers. The California requirement had been in force for more than two years before the regulator found non-compliance at the two sites.

Failure Pattern: work was measured and disciplined by an automated productivity count, and the requirement to tell workers what the count expected of them was still not being met years after a law required it.

Governance Principle: where an automated measure of a person’s work can lead to discipline, the standard being applied should be disclosed to that person in writing, and each consequential decision should be recorded in a way that shows what a human actually decided.

The reference implementation of ConductRecord and MetricRecord is open source. It lives at github.com/saffronandindia/headlights-oss, Apache 2.0 licensed and free to install. The repository is public now.

Sources

The mailing list

Fresh incident reports every week. One email to match.

We add new incidents to the library regularly, and send a single short email each week with what's new. The library stays free and open; this is just how you keep up with it.

No tracking. Unsubscribe in one click.

The record

An auditable system would have produced a signed, tamper-evident record the moment this happened: what the system did, the version that did it, the basis it acted on, and the action taken, and Amazon.com, Inc. could have produced it on demand.

This is the record the system as deployed did not produce in a signed, auditable form.

What this teaches
Capture what happened when it happens
What the system did, the version that did it, the basis it acted on, and the action taken, recorded at the moment, not reconstructed after.
Sign it, so no one has to trust the record-keeper
A tamper-evident entry. Edit it later and the signature breaks. The record does not ask for the benefit of the doubt.
Make it verifiable by anyone
A court, a regulator, a customer's lawyer can check the record themselves, without taking the company, or us, at our word.

Headlights summarises publicly reported AI incidents. All summaries are independently written, attributed to their original sources, and intended for research and educational purposes. Allegations are identified as such until established through official findings.

This report is based on the sources listed above and reflects information available at the time of review; later developments may not be captured. Where a person is described as charged with or alleged to have done something, that allegation is unproven unless a conviction or a court or regulatory finding is stated. Headlights publishes journalism and commentary, not legal advice.

Want to write back?

Direct to my inbox.

ellie@useheadlights.com →