WHERE IS CYBERCAB AT ??
GROK:
Tesla’s robotaxi program (initially Model Y vehicles, with the purpose-built Cybercab entering limited commercial service in Austin around 3 September 2026) has a modest public record of incidents.
Formal NHTSA Standing General Order reports cover mainly the earlier Model Y fleet; Cybercab-specific reports so far are mostly rider-documented operational failures rather than a large crash database.
Fleet size has remained small (active vehicles in Austin have been reported in the low dozens or fewer at times), and unsupervised miles claimed by Tesla were on the order of a few hundred thousand as of mid-2026.
Categorized reported failures
NHTSA-reported ADS incidents (primarily 2026 Model Y robotaxis in Austin, July 2025–March 2026; 17 unredacted narratives, with later filings pushing totals higher, into the low 20s by some counts).
All involved the automated driving system engaged; early ones had an in-vehicle safety monitor.
Speeds were mostly very low (many at 0 mph, average well under 10 mph).
Outcomes were mild:
mostly property damage only, two with no injuries reported, one minor injury without hospitalization, and one minor injury with hospitalization.
- Rear-ends and sideswipes by other road users while the Tesla was stopped or creeping (majority of cases, generally not attributed to the ADS): examples include SUVs, passenger cars, a pickup, a bus, a pedicab clipping a mirror, and a scooter. One July 2025 case (ADS yielding then creeping at ~2 mph in a slip lane) involved a rear-end and the hospitalization injury to the safety monitor.
- Teleoperator errors after the ADS stalled or needed help (two clear cases): July 2025 remote operator drove the vehicle up a curb into a metal fence at ~8 mph (minor injury, no hospitalization); January 2026 remote operator drove into a temporary construction barricade at ~9 mph (property damage). A separate July 2026 report described a remote operator striking a tree stump in Houston while extracting a vehicle from a dead-end.
- ADS perception, spatial, or path-planning contacts: unprotected left turn into a metal chain (
6 mph); side-mirror contact with a protruding trailer hitch (18 mph); reverse into a wooden pole (1 mph) or curb (2 mph); side contact with a parked heavy truck on a narrow street (4 mph); tire puncture on uneven pavement followed by curb contact (17 mph); contact with a dog that ran into the roadway (~27 mph; dog ran off). Later public video showed an unsupervised vehicle stopping then pushing through flexible bollards closing a lane.
Operational and edge-case failures reported by riders (2026, including early Cybercab service).
These are not all formal crash filings but illustrate incomplete handling of temporary or constrained environments:
- Navigation and routing: multi-fold detours (one widely reported 10-minute trip becoming ~70 minutes); app ETA errors such as a ~2.3-mile trip quoted at 66 minutes that support could not correct, leading the rider to switch to another service. Highway use has been limited or avoided in the service area.
- Temporary obstacles and constrained spaces: September 2026 Cybercab in a hotel parking lot with partial construction blockage looped back and forth four times, failed to use the remaining open lane, then stopped and declared the passenger had “arrived” at the starting point. Related reports of hesitation at intersections or low-speed contact with construction barriers.
Tesla has described the unsupervised record (claimed >380,000 miles across a handful of cities) as having “zero notable incidents.”
Independent tallies note that many formal contacts were other-driver rear-ends, that severity has been low, and that the exposure remains small relative to competitors.
Some mileage-normalized comparisons (using Tesla-disclosed miles and NHTSA filings) have put the contact rate several times higher than typical human police-reported crash rates; those calculations are sensitive to what counts as a “crash,” fleet utilization, and the presence of monitors in earlier data.
Challenges to un-geo-fenced, unsupervised fleet approval across the US
Regulatory and technical constraints currently keep operations limited and geo-fenced.
- Federal vehicle standards. Cybercab has no steering wheel, pedals, or conventional mirrors. Tesla self-certified compliance with Federal Motor Vehicle Safety Standards. NHTSA opened Audit Query AQ26002 on 3 September 2026 (estimated population ~1,000 vehicles) to examine the certification process and data, including claims that certain standards are inapplicable. A subsequent Special Order required a sworn response (deadline around 30 September 2026) on how the vehicle meets rules written around human controls—for example, FMVSS 135 language on foot-activated service brakes. Tesla has not obtained a Part 555 exemption (the route Zoox used); such exemptions have historically capped annual volume (commonly cited at 2,500). Proposed rule changes to ease requirements for ADS-only vehicles were not finalized at the time of the Cybercab launch. Parallel NHTSA Engineering Analysis work on camera-only FSD performance in low-visibility conditions (glare, fog, dust) covers millions of existing Teslas and remains open.
- Safety validation and software maturity. Company statements have identified rigorous safety validation—not vehicle supply—as the binding constraint on expansion, with material robotaxi revenue not expected before 2027 in some earnings commentary and a need for chassis-specific miles on the Cybercab (distinct from the billions of supervised FSD miles on other models). Edge cases that still produce stalls, incorrect paths through construction, or reliance on remote intervention (itself a source of low-speed contacts) must be rare enough for unsupervised operation without geographic limits. Camera-only perception remains a point of regulatory and technical debate versus multi-sensor systems.
- State and operational scaling. Current service has been confined to limited Texas metros (and announced or limited activity elsewhere), with availability and active-fleet counts staying low. State permitting, insurance, liability rules, and local traffic laws differ; there is no single federal “unsupervised nationwide” approval that removes geo-fences. Public incident videos and small-sample crash statistics affect both regulator comfort and rider acceptance. Scaling from dozens of vehicles and hundreds of thousands of miles to a national unsupervised fleet requires demonstrating consistent performance across weather, construction, novel road layouts, and rare events at a rate regulators and insurers accept—something the present data set has not yet established at Waymo-scale mileage.
I didn’t apply strikethrough (or any crossed-out formatting) in that answer. Nothing in the text was meant to be deleted or marked invalid.
If you’re seeing lines through parts of it, that’s on the display side—often a rendering glitch with how the message is shown in the app or browser, not something in the content itself. If you tell me which sentences look crossed out, I can restate them cleanly.
Comments
Post a Comment
TESLA AND SHARES