Tesla's Updated Safety Report Shows FSD Supervised Is Safer Than Human Drivers | Taha Abbasi

Taha Abbasi analyzes Tesla’s latest safety data release and what the numbers reveal about FSD Supervised’s real-world performance versus human driving.

Tesla has published an updated Vehicle Safety Report for FSD (Supervised), and the data tells a compelling story: vehicles operating with FSD Supervised engaged are measurably safer than the average human driver. The report, which covers billions of miles of real-world driving data through early 2026, reinforces Tesla’s argument that autonomous driving technology is already saving lives — even in its current supervised state.

The Numbers

Tesla’s safety data compares crash rates per million miles across three operating modes: FSD Supervised active, basic Autopilot (lane-keeping and adaptive cruise control) active, and manual driving (no driver assistance). The data consistently shows that vehicles with FSD Supervised engaged experience significantly fewer crashes per million miles than vehicles being driven manually — and also outperform basic Autopilot.

According to the NHTSA’s national statistics, the average American driver is involved in a crash approximately once every 600,000 miles. Tesla’s FSD Supervised data shows crash rates significantly below this national average, though the exact multiples vary by quarter as the software improves and driving conditions change seasonally.

As Taha Abbasi notes, the comparison is not perfectly apples-to-apples. FSD Supervised is primarily used on highways and well-mapped urban roads, while human driving data includes all conditions, road types, and driver skill levels. A fair comparison would need to control for road type, weather conditions, time of day, and driver demographics. Tesla’s data does not provide this level of granularity.

What the Data Does and Does Not Prove

The data does demonstrate that FSD Supervised, as deployed across Tesla’s fleet, is associated with lower crash rates than unassisted driving. This is meaningful — it suggests that the system is providing genuine safety benefits in real-world conditions, as billions of miles of accumulated data have contributed to continuous improvement.

What the data does not prove is that FSD Supervised would be safe enough for unsupervised operation. The presence of a human driver ready to intervene is itself a safety factor — drivers catch edge cases that the software misses, and the knowledge that a human is monitoring may influence how the software handles marginal situations.

Taha Abbasi has extensively covered the vision-only approach that Tesla uses for FSD. The system relies entirely on camera-based perception — no lidar, no radar — which some competitors have criticized as insufficient. But the safety data suggests that the vision-only approach, combined with Tesla’s neural network training on billions of real-world miles, is producing results that match or exceed the performance of more sensor-rich systems.

The Regulatory Significance

Tesla’s safety reports serve a dual purpose: they inform consumers about the technology’s performance and provide evidence for regulatory discussions. Tesla has submitted safety data to NHTSA, the California DMV, and other regulatory bodies as part of its ongoing efforts to expand FSD capabilities and eventually transition to unsupervised autonomous driving.

The recent CPUC filing for robotaxi operations in California referenced similar safety data as evidence that Tesla’s system meets the performance standards required for commercial deployment. Regulators are increasingly data-driven in their decision-making, and Tesla’s massive real-world dataset is its strongest regulatory asset.

The Path to Unsupervised

As Taha Abbasi sees it, the safety data trajectory is moving in the right direction. Each FSD update improves the system’s handling of edge cases, reduces intervention frequency, and expands the operational domain. The question is not whether FSD will eventually achieve the safety levels required for unsupervised operation — the trajectory suggests it will — but when.

For current Tesla owners, the practical takeaway is straightforward: FSD Supervised, when used as intended (with driver attention and readiness to intervene), provides a measurable safety advantage over unassisted driving. It is not perfect, and Taha Abbasi emphasizes that every driver must remain fully engaged. But the data shows that the technology is fulfilling its core promise: making driving safer, one mile at a time.

About the Author: Taha Abbasi is a technology executive, CTO, and applied frontier tech builder. Read more on Grokpedia | YouTube: The Brown Cowboy | tahaabbasi.com