Tesla's 'self-driving' software fails at train crossings (www.nbcnews.com)

Multiple Tesla owners have reported that Full Self-Driving (FSD) software mishandles railroad crossings—failing to recognize flashing lights, lowering gate arms, or oncoming trains—forcing drivers to intervene to avoid collisions. NBC News reviewed videos from a North Texas owner whose 2025 Model Y (HW4, running FSD 13.2.9) nearly drove through a closing crossing and replicated the failure on a follow-up visit; the outlet also found numerous other clips and dozens of forum posts documenting similar failures dating back to mid‑2023. At least one Tesla in FSD mode was later struck by a freight train after entering tracks. The National Highway Traffic Safety Administration (NHTSA) has contacted Tesla and is already investigating related FSD visibility issues. Technically, experts point to Tesla’s camera‑based, neural‑network approach and a “black‑box” model architecture that makes failure modes hard to explain and correct; engineers likely underrepresented railroad crossings in training data, producing inconsistent behavior. The incidents underscore real safety risks for a Level‑2 system that Tesla markets aggressively while still requiring active human supervision. Beyond immediate crash danger, the pattern raises regulatory scrutiny, questions about training-data completeness and model interpretability, and tensions between Tesla’s autonomy claims (and robotaxi plans) and the demonstrated limitations of current FSD deployments.