The Machine Record: a sourced register of incidents involving commercial robots and autonomous vehicles, followed from first report to legal or regulatory outcome
Year: 2026
Type: preprint
Abstract: Data descriptor, preprint version 1 (26 October 2026). Public records of harm caused by commercial robots are fragmented across news coverage, regulator filings and court dockets, and existing AI incident databases record first reports rather than outcomes. The Machine Record is an open register of incidents involving commercial physical robots (autonomous vehicles and robotaxis, sidewalk delivery robots, humanoids, industrial arms and cobots, warehouse, surgical, consumer and service robots) in which each case is followed from the first report through any investigation, regulator action, recall, fine, lawsuit, settlement or criminal charge. The release described here holds 255 verified cases dated 1979 to 2026, 302 timeline events, 640 sources across 330 domains, of which 592 are archived, and 1,551 crash filings made by automated driving system operators under the United States National Highway Traffic Safety Administration Standing General Order, linked to news cases where the filing and the case describe the same event. Cases enter through a published, versioned set of publication gates implemented as an open-source package (tmr-gates), requiring at least two independent sources, a dated event and a named machine, with stricter rules for deaths and serious injuries. The data are distributed under CC BY 4.0 as CSV and JSON with a codebook, annotation guidelines and a limitations statement, mirrored daily on GitHub and archived quarterly on Zenodo. Data: https://doi.org/10.5281/zenodo.22925531. Register: https://themachinerecord.com. Publication gates: https://github.com/vzolot/tmr-gates.
Author: Zolotukhin, Volodymyr
Language: English
Subfield: Law
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7223704533