Tesla Engineers Challenge Autopilot Safety Marketing Claims
Current and former Tesla engineers have challenged Autopilot safety claims, alleging internal crash data uses methodology specifically designed to make comparisons with human driving appear more favorable. Tesla publishes quarterly reports showing Autopilot has fewer crashes per mile, but engineers allege these statistics are misleading because Autopilot is disproportionately used on highways in favorable conditions.
Methodology Concerns
Tesla compares Autopilot crash rates against NHTSA averages for all driving. Engineers argue this is misleading since Autopilot operates primarily on divided highways in clear weather, the safest environments. The NHTSA average includes urban intersections, rural roads, and adverse conditions. When controlling for road type and weather, the safety advantage narrowed significantly or disappeared.
Disengagement Event Classification
Engineers describe disengagement events before crashes not counting in statistics because the system was technically not engaged at impact, even when disengagement and crash were the same event sequence. Phantom braking events, where Autopilot brakes for non-existent obstacles, are classified as driver interventions rather than system failures.
Regulatory Implications
NHTSA has opened multiple investigations and issued recalls. However, Tesla marketing continues using quarterly reports as safety evidence. Engineers argue regulators lack technical expertise to evaluate methodology independently.
Key Findings
- Tesla Autopilot safety comparisons use methodology comparing highway driving to all-road-type averages
- Autopilot disengagement events immediately before crashes are excluded from crash statistics
- Phantom braking incidents classified as driver interventions rather than system failures
Timeline
NHTSA opens formal Autopilot investigation
Tesla recalls 2 million vehicles for monitoring deficiencies
Former engineers submit methodology critique to NHTSA
NHTSA proposes new evaluation framework