Tesla AI trainers distrust Full Self-Driving safety claims, Reuters investigation finds

A Reuters investigation cited in the research found that Tesla’s own AI data labelers and a former self-driving engineer expressed significant distrust in the company’s Full Self-Driving system and its publicly promoted safety statistics.

The report was based on interviews with former employees and traffic-safety researchers. It found that safety claims around Full Self-Driving relied on flawed methodology, according to the research summary.

Data labelers described regularly observing failures by the system in basic driving tasks, including responding to emergency vehicles, navigating construction zones and recognizing pedestrians. Some former employees said they would not trust the technology to drive them.

The findings add scrutiny to Tesla’s presentation of Full Self-Driving as the company continues to promote the technology’s capabilities. The investigation focuses not only on system performance but also on whether the safety statistics used to support public claims accurately reflect real-world risks.