Tesla reported 207 crashes linked to its driver-assistance systems in May 2026, the largest single-month total recorded in public filings to the National Highway Traffic Safety Administration (NHTSA). The filings are part of NHTSA’s Standing General Order (SGO) on advanced driver-assistance systems (ADAS), but the company has redacted most of the information that would explain what happened in these incidents.
NHTSA reporting framework and what is captured
The NHTSA SGO requires automakers to report crashes in which an ADAS was engaged within roughly 30 seconds of impact and the crash meets a severity threshold — defined as airbag deployment, a tow-away, an injury, a fatality, or a struck pedestrian or cyclist. Manufacturers submit telematics-based crash records under that order, and those records form the dataset used to track ADAS-related incidents across the industry.
Record month and an accelerating trend
Electrek merged the latest NHTSA ADAS export with the archive back to 2019 and deduplicated to the most recent version of each report. That analysis shows Tesla logged 207 crashes in May 2026, a single-month total larger than all of 2021 combined (157). Across the file Tesla now accounts for 3,763 unique driver-assist crash reports, representing about 85% of the industry’s ADAS records in the dataset.
The broader trend shows rising counts year over year. Tesla was the first company in the dataset to exceed 1,000 driver-assist crash reports in a single year, with 1,043 in 2025. For the first six months of 2026 the file already contains 826 Tesla-related incidents, putting the year on pace to outpace 2025 substantially. Comparing identical January-to-June windows across years, the counts run 180, 261, 269, 476, and 826 — a 73% increase year over year off an already larger base.
Why raw counts don’t tell the full story
Raw incident counts increase as a fleet grows and as drivers run ADAS for more miles. Tesla has sold more vehicles and expanded Full Self-Driving (FSD) subscriptions, meaning more engaged miles could produce more crashes even if the technology is safer per mile. But outside analysts cannot calculate a crash rate because Tesla does not publish independently auditable miles-driven figures for Autopilot and FSD. The company’s quarterly safety reporting has also faced criticism, including from people who have worked as AI trainers for Tesla.
Because the denominator — engaged miles — is held only by Tesla, the public dataset cannot determine whether the per-mile crash rate is improving, worsening, or steady.
Extensive redactions limit transparency
One striking finding is how little detail Tesla has made public. Tesla redacted the narrative text in 99.9% of its 3,763 crash reports, citing confidential business information. The severity field in most entries is listed as “Unknown” because the underlying telematics descriptions are redacted. Tesla has also redacted the software-version field, which prevents separating incidents involving Autopilot from those involving FSD in the public file.
By contrast, in the same export automakers such as GM, Ford, Honda, and Toyota redacted almost none of their narratives. Electrek notes that Tesla is the only high-volume reporter that systematically blacks out these details.
Tesla has previously told regulators it would suffer financial harm if the information were public. The company is also subject to a separate NHTSA probe examining how it reports ADAS-related crashes.
Undercounting means the May total is a floor
Reporting lag affects counts in recent months. When the latest export was added it revised a prior “final” May figure upward by 139 crashes. That revision implies the 207 reported for May 2026 should be considered a minimum estimate rather than a definitive total.
Context and editorial perspective
Electrek’s coverage emphasizes that FSD has been improving — the outlet reports hands-on testing of v14 and calls it the most capable Level 2 system available to consumers. But the editorial argues that improvements in capability can drive complacency among drivers, especially when marketing positions the system as “Full Self-Driving.” According to that perspective, drivers treating Level 2 assistance as an unsupervised service can lead to distractions and misuses that increase crash counts, even as the software advances.
Electrek also points out a mismatch between Tesla’s public messaging and actions: consumer FSD remains a Level 2 assist that requires driver supervision, while the company’s promised robotaxi service has not been scaled — the fleet remains small. Those actions, the outlet contends, suggest Tesla is not yet deploying an unsupervised system at scale.
What this means going forward
The NHTSA SGO records show that crashes tied to Tesla’s driver-assist systems are rising in absolute terms and that the company dominates the public dataset. But without independently verifiable mileage figures and with nearly universal redaction of crash narratives, outside parties cannot fully assess the systems’ safety on a per-mile basis or understand the circumstances that led to most incidents. That gap in transparency is likely to remain a central point of regulatory and public scrutiny as ADAS use and capabilities expand.
Source: Electrek