Tesla Is Dropping the Dream of Human-Free Self-Driving Cars (www.vice.com)

Tesla has quietly rebranded its controversial "Full Self-Driving" (FSD) product as "Full Self-Driving (Supervised)" and updated fine print to make clear cars built from 2016–2023 require human supervision and are not autonomous by strict definitions. The company has sold FSD options for up to ~$15,000 and is also pursuing a massive executive payout tied to milestones including "10 million active FSD subscriptions." Observers (notably Electrek) flagged that Tesla’s public definitions are broad enough to encompass many driver-assist features, allowing the company to preserve the FSD label on paper while delivering a supervised, driver-in-loop product in practice. For the AI/ML community this underlines a hard truth: deployed systems that look like autonomy are often advanced ADAS, not true SAE Level 4/5 autonomy. Technically, Tesla’s stack remains a supervised vision and control system that relies on a human fallback, rather than a fully validated edge-case-handling autonomous agent. That has implications for dataset coverage, validation, safety claims, liability and regulatory scrutiny—especially when commercial incentives (subscription counts, executive pay) can motivate optimistic labeling. The episode highlights the gap between marketing and measurable autonomy, and stresses the importance of clear definitions, rigorous testing, and transparent performance metrics for any system claiming to be "self-driving."