Tesla Autopilot Disengages Moments Before Crash in Test, Revealing Shady Practice

A recent crash test video by Mark Rober, a famous engineering Youtuber, has brought to light a potentially problematic behavior of Tesla’s Autopilot system. The video, which compared Tesla’s camera-based ADAS with a lidar-based system, showed a Tesla Model Y on Autopilot failing to stop for a simulated road obstacle. Initially, Tesla fans and investors, particularly a group called ‘Rebellionaires,’ attempted to discredit the video, claiming that Autopilot was not even active during the test and accusing the author of publishing fake news. However, the video clearly shows Autopilot being engaged. More critically, it revealed that Autopilot disengages less than a second before the crash. This detail, initially missed by the author, was highlighted by the Tesla investors’ attacks. This behavior, where Autopilot disengages just before an impact, has been previously documented. NHTSA’s investigation into Tesla Autopilot crashes involving emergency vehicles found that Autopilot would disengage within one second of impact. Critics argue this could be a deliberate attempt by Tesla to create legal confusion and avoid liability in accidents. Furthermore, the article points out Tesla’s history of avoiding transparency with ADAS data, including withholding data from the California DMV and abusing NHTSA’s crash redaction policies, further fueling concerns about the company’s handling of Autopilot-related safety issues.