2026-06-17 👁️ — views

AV Safety & Regulatory Scoreboard — Tesla vs Waymo: Who Has the Clearer Path to Scale?

California DMV reports, NHTSA crash data, and state permit maps reveal who leads in autonomous vehicle safety metrics and regulatory readiness as of mid-2026.

Why regulatory data is the real scoreboard

Weekly ride counts and supervised-mile totals are marketing metrics. The deeper question for any autonomous vehicle program is: can it legally operate without a human in the seat, and how safely has it done so in publicly verified data? Two mandatory reporting frameworks answer that question — California DMV AV program reports and NHTSA Standing General Order crash data. Together they form the only publicly available, government-collected safety benchmark for US AV programs as of mid-2026.

Section 1 — California DMV AV Reports: the public benchmark

California is the strictest and most data-rich AV testing state. The DMV requires all companies testing autonomous vehicles on public roads to file:

1. Annual disengagement reports — how often a human safety driver had to take over per 1,000 miles driven
2. Collision reports — every crash or contact involving an AV test vehicle on public roads

Important caveat on Tesla: Tesla does NOT file California DMV AV disengagement or collision reports. FSD (Full Self-Driving) is classified by Tesla as a driver-assistance system requiring constant human supervision, not a SAE Level 4 autonomous test program. As a result, Tesla’s FSD data does not appear in the California DMV AV reporting system. Tesla’s safety metrics come instead from its own voluntary disclosures and from NHTSA crash reports (Section 2).

California AV Program Comparison Table (most recent annual reporting period available through mid-2026)

What Cruise’s suspension teaches: In October 2023, Cruise — GM’s robotaxi subsidiary — had one of its vehicles involved in a collision that dragged a pedestrian. The California DMV suspended Cruise’s driverless operating permit within weeks. Cruise halted all US operations. GM subsequently wound down the robotaxi division. The incident demonstrated that a single high-profile safety event can terminate a program, regardless of prior miles accumulated. Regulators will not grandfather past performance against present safety failures.

Waymo’s disengagement trend: Waymo’s annual disengagement rate has declined steadily across its filings, from above 1.0 per 1,000 miles in earlier years to consistently below 0.5 in recent reports — with some recent periods showing rates below 0.1 per 1,000 miles in mature operating zones. No other CA filer with comparable commercial scale has matched this figure.

Section 2 — NHTSA Standing General Order data: volume vs rate

In June 2021, NHTSA issued a Standing General Order (SGO) requiring manufacturers and operators of vehicles with SAE Level 2 and above automation to report crashes within 30 days. This is the broadest mandatory AV crash-reporting framework in the US.

Key findings from SGO data (through mid-2026)

Tesla leads in total incidents reported — by a large margin. This is primarily a volume effect: Tesla has more than 6 million FSD-capable vehicles on US roads, all subject to reporting whenever FSD is engaged at the time of a crash. Even a very low crash rate per mile produces a large absolute number when multiplied by billions of supervised miles.

Waymo reports far fewer absolute incidents. Waymo’s commercial fleet is roughly 1,000–1,500 purpose-built vehicles. At that scale, even an identical crash rate would produce fewer total reports.

The critical distinction is rate, not count. The SGO data alone does not publish per-mile crash rates with full statistical rigor — NHTSA has cautioned against naive comparisons between operators. A program with more miles, more vehicles, and more types of road exposure (highway, city, rural) will generate more reports. Responsible analysis requires normalizing by miles driven or vehicle-hours exposed.

What the data does show clearly:

• Tesla’s FSD incidents are nearly all supervised (human driver present, hands legally on or near wheel)
• Waymo incidents occur during fully driverless commercial operation
• The two programs are fundamentally different in exposure type, making direct rate comparison complex

Section 3 — State-by-state AV regulatory map

The ability to operate commercially without a human safety driver is the legal threshold that separates a research program from a scalable business. States vary dramatically in how they regulate this.

Driverless Commercial Operation Permit Matrix (as of mid-2026)

Why Tesla chose Austin, not San Francisco: Texas requires no state-level driverless permit — operators self-certify under the 2017 Texas AV statute. California requires a DMV application, safety driver removal petition, and ongoing reporting. For Tesla’s first robotaxi launch without safety drivers, Texas offered the fastest path to revenue. Waymo operates in both states but built its California footprint over years of DMV permit compliance.

Section 4 — Regulatory milestones needed for Tesla to scale nationally

Tesla’s Robotaxi ambition — especially the Cybercab, a two-seat vehicle with no steering wheel or pedals — faces a layered regulatory stack that Waymo has largely already cleared.

Milestone 1: FMVSS exemption for the Cybercab

Federal Motor Vehicle Safety Standards (FMVSS) require vehicles sold in the US to have a steering wheel, brake pedal, and accelerator. A vehicle with none of these — like the Cybercab — cannot be manufactured for road use without an NHTSA exemption under 49 U.S.C. § 30113.

NHTSA has granted exemptions before (Waymo’s sixth-generation vehicle received a partial exemption), but the process involves a public comment period, safety justification, and volume limits. Tesla filed for a Cybercab FMVSS exemption in 2025. NHTSA review timelines have historically ranged from 12 to 36 months for complex petitions. Estimated resolution: late 2026 to 2027.

Milestone 2: State-by-state driverless permits (Texas first, then others)

Even in permissive Texas, Tesla must demonstrate operational safety at scale before expanding. The progression follows a predictable pattern: geofenced pilot → permit application or self-certification → expand geofence → new city. For California specifically, Tesla would need to apply for a CA DMV driverless deployment permit — a process that requires documented test miles, safety case submission, and regulatory review. Estimated CA timeline: not before 2027–2028, assuming FSD disengagement rate continues improving.

Milestone 3: Remote monitoring and minimum-risk condition compliance

NHTSA’s forthcoming AV policy framework (successor to AV 3.0/4.0 guidance) is expected to require fully driverless vehicles to demonstrate minimum-risk condition (MRC) capability — the ability to safely pull over and stop if the system fails — and 24/7 remote monitoring capability. Tesla is building a remote-assistance infrastructure, but it is not yet publicly documented at commercial scale. Estimated framework clarity: 2026–2027.

Milestone 4: Federal type approval for vehicles without driver monitoring systems

Current FMVSS also requires driver monitoring systems (DMS) in many new-vehicle categories. A vehicle intended to operate without any occupant capable of driving needs a new federal regulatory category. NHTSA proposed a new rule framework in 2023; final rules have not yet been issued as of mid-2026. Estimated finalization: 2027.

Tesla regulatory milestone timeline (estimates)

Section 5 — Safety verdict: who has the clearer regulatory path

Waymo holds the clearer regulatory path today. It has already cleared the hardest gates:

• Full driverless commercial permits in California, Arizona, and Texas
• Years of DMV disengagement report filings with a declining and low disengagement rate
• Proven operational track record at commercial scale (150,000+ weekly paid rides)
• A purpose-built vehicle (Gen 6) designed from the ground up to meet driverless standards
• NHTSA coordination on its sixth-generation vehicle design

Tesla’s path is real but longer. FSD’s supervised miles are a genuine data asset — no other AV program has accumulated billions of miles of road data from a consumer fleet. The question is whether that data advantage translates to regulatory approval fast enough to matter competitively. The Cybercab’s FMVSS exemption, state-by-state permit progression, and the absence of any CA DMV AV program filing all add time to Tesla’s national scale timeline.

The Cruise lesson applies to everyone. A single serious safety incident can trigger permit suspension. Both Waymo and Tesla are operating under intense public scrutiny. Waymo’s operational maturity reduces (but does not eliminate) that risk. Tesla’s supervised-only status means that FSD incidents are legally the driver’s responsibility — a regulatory buffer that disappears the moment Tesla applies for driverless permits.

Bottom line for mid-2026: Waymo is in commercial operation across four US cities with regulatory approval in hand. Tesla has launched a supervised robotaxi pilot in Austin and faces a 12-to-36-month stack of regulatory milestones before it can operate Cybercabs at scale without safety drivers. The data gap in publicly verifiable safety metrics — Waymo files; Tesla does not file under the same framework — means independent verification of Tesla’s safety record at the fleet level remains limited to NHTSA SGO reports alone.

The regulatory finish line is not a single approval. It is a continuous compliance relationship with regulators across every city of operation. Waymo has built that relationship over seven years. Tesla is building it now.

Sources

• California DMV AV testing reports ↗
• NHTSA Standing General Order crash data ↗
• NHTSA AV crash reporting dashboard ↗
• California DMV AV disengagement reports ↗
• Tesla NHTSA special order responses — public docket ↗

Tags