Tesla FSD data labelers refuse to ride in their own system, citing safety concerns

At a glance:

• Seven of nine former Tesla data labelers told Reuters they would not ride in a vehicle using Full Self-Driving (FSD) technology.
• Data specialists reported routine speeding and system failures in FSD footage, which were deprioritized by engineers.
• Tesla expanded FSD availability to China amid ongoing safety concerns and Musk's unmet promises for fully autonomous driving.

Inside the FSD Training Process

Former Tesla data labelers played a critical role in shaping the performance of Full Self-Driving (FSD) by reviewing hours of driving footage and flagging errors. These workers had direct access to terabytes of proprietary data, witnessing firsthand how the system navigated real-world roads. However, their proximity to the technology did not translate into confidence. Seven out of nine labelers interviewed by Reuters said they would not ride in a Tesla operating on FSD, with one stating they would avoid robotaxis "if you f**king paid me." A former self-driving engineer echoed these concerns, saying, "We have all seen it fail," and questioning CEO Elon Musk's claims of "safe unsupervised" driving readiness.

The labelers' responsibilities included training the AI to avoid past mistakes, yet they observed recurring issues such as vehicles exceeding speed limits while in FSD mode. At least five interviewees noted that routine speeding—which impacts every drive—was treated as a low priority compared to edge-case scenarios like unusual road layouts or lighting. This prioritization gap highlights a disconnect between Tesla's public messaging and internal development focus.

Musk's Promises vs. Reality

Tesla has long marketed FSD as a stepping stone to fully autonomous vehicles, but Musk's timeline for achieving this goal has repeatedly slipped. Since 2016, the CEO has promised Level 4 autonomy, yet the current FSD (Supervised) system remains classified as Level 2, requiring constant driver attention. Recent regulatory filings and marketing materials emphasize safety statistics, but the former engineer disputed these claims, suggesting they are misleading. The company's robotaxi trials in Austin, Texas, operate within geofenced areas with remote safety drivers, underscoring the gap between Musk's vision and current capabilities.

Safety Incidents and Industry Comparisons

Recent months have seen multiple FSD-related accidents, including vehicles driving into lakes, off bridges, and into oncoming train paths. While these incidents made headlines, the data labelers' testimony implies a broader catalog of failures in internal footage. Tesla's approach to autonomy—relying solely on cameras rather than multi-sensor fusion—differs sharply from competitors like Waymo, which uses lidar and radar. Even Waymo's systems, however, have shown vulnerabilities, such as recent shutdowns during flooding, illustrating that no autonomous technology is immune to real-world challenges.

Implications for Tesla and Consumers

The reluctance of FSD trainers to use the system raises fundamental questions about its reliability. If the individuals closest to the data lack trust, it complicates Tesla's narrative for consumers and regulators. The expansion of FSD to China, where regulatory oversight is stringent, adds further scrutiny. Tesla has not responded to Reuters' inquiries, leaving unresolved concerns about transparency and safety protocols.

The investigation underscores the risks of conflating marketing claims with technical reality. As autonomous driving becomes a cornerstone of automotive innovation, the credibility of developers and their willingness to use their own products may become a litmus test for public adoption.

What's Next for FSD Development

Tesla's robotaxi trials in Austin will continue to serve as a testing ground for unsupervised autonomy, though the geofenced limitations suggest significant hurdles remain. The company's future regulatory filings may face increased scrutiny from agencies questioning the validity of safety claims. Meanwhile, the data labelers' experiences highlight the need for clearer communication between internal teams and leadership about real-world performance gaps.

For consumers, the message is cautious: despite Tesla's promotional push, FSD remains a supervised system with unresolved flaws. The industry's broader shift toward transparency and accountability may pressure Tesla to address these concerns more openly.