Safety of Tesla Systems Questioned by Independent Investigation
A Reuters investigation has revealed serious flaws in Tesla‘s self-driving software and exposed methodological errors in its official road accident statistics. The report based its findings on direct interviews with former employees involved in artificial intelligence training and on analysis by transportation safety experts. The collected testimonies contradict the automaker’s public claims about the reliability of its driverless vehicles and their imminent international rollout.
The journalistic revelation comes at a critical moment for Elon Musk‘s company, given its promises to commercialize robotaxis in the near future. Experts warn that the discrepancy between corporate data and operational reality hinders the company’s ability to obtain the necessary government permits for public passenger transport. Documenting these technical inconsistencies places the American firm’s autonomous mobility strategy under regulatory scrutiny.
Distrust in the AI training team
The reporting included in-depth interviews with nine former data labelers who annotated the visual data used to train Tesla’s Full Self-Driving system. Seven of the nine specialists explicitly stated that they do not trust this software for personal use due to the constant errors they witnessed daily in test recordings.
The labelers highlighted recurring AI failures, noting that the system struggled to identify stationary road hazards, construction zones, and moving emergency vehicles.
The technical staff also indicated the existence of strong internal pressure within the development teams to prioritize route flow over safety margins. Programmers were required to adjust algorithm parameters to prevent sudden braking or hesitant vehicle behavior in complex urban environments. This management approach weakened risk control protocols in the pre-validation of software updates.
Inconsistencies in accident statistics
The investigation revealed that Tesla artificially inflates its road safety records by a factor of up to three compared to the reality on the roads. The analysis determined that the corporation’s methodology is biased, as it compares serious crashes involving its new models only against broad federal statistics that include minor collisions across the entire national vehicle fleet.
Ten of the eleven independent traffic safety researchers consulted by the agency rated the company’s reporting metrics as deceptive marketing.
The written report demonstrated that the company’s reports omit incidents occurring on secondary roads or those that do not trigger the vehicle’s airbags. This systematic exclusion of data obscures the true frequency of minor accidents and distorts the calculation of collision-free mileage. Traffic authorities are currently evaluating these methodological findings as part of the technical audits they conduct on the American firm’s prototypes.