Tesla FSD AI trainers report safety concerns to Reuters
A Reuters investigation published on 28 May 2026 reported that nine former Tesla data annotators and an ex-engineer expressed safety concerns about Full Self-Driving, stating that seven of nine trainers refused to ride in vehicles using the system they helped develop. The report also documented that ten of eleven road safety researchers considered Tesla's published safety statistics misleading.
What happened
Nine former Tesla data annotators, one ex-engineer, and eleven road safety researchers provided testimony to Reuters in an investigation published 28 May 2026. Seven of nine annotators said they refused to ride in vehicles piloted by the Full Self-Driving software they had helped train.
Annotators described sequences where the system struck animals without braking, came close to children at pedestrian crossings, and exceeded speed limits by 30–50 km/h in what one called 'Mad Max' mode. An ex-engineer stated 'Do not trust Elon on this.'
Reported impact
Ten of eleven road safety researchers characterized Tesla's published safety statistics as 'misleading marketing' rather than rigorous analysis. The investigation identified three methodological biases in Tesla's claim that FSD is 'up to ten times safer' than human drivers.
Tesla's comparison method counted only airbag-deployment accidents in its vehicles against the federal U.S. rate including all tow-away accidents; applied different fleet age demographics; and used a five-second responsibility window versus the thirty-second federal standard. Correcting these biases reduced the claimed safety margin by a factor of three.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.