Teslarati reported on 18 May 2026 that Tesla's internal Robotaxi app version 26.4.5 carries a substantial amount of new code aimed at remote operations, safety protocols and autonomous ride-hailing flow (Tesla making sweeping improvements to Robotaxi app). This is Tesla's operator-facing tool for the fleet — not the consumer Tesla App — so the changes describe how Tesla manages the cars currently driving paid passengers in Austin, Houston and the Dallas area. Read in context, "sweeping improvements" here means tightening control over a fleet that is still on the order of dozens of vehicles, not thousands.
A high-security remote kill-switch
The most striking addition flagged in the code is a high-security command that can freeze a vehicle's ability to drive. According to Teslarati's reading of the app strings, the command pulls a car out of the Robotaxi fleet for any reason chosen by the operator and prevents the vehicle from being put into gear — even when somebody is physically present with valid keys. It is a remote disable that cannot be overridden at the car itself.
For a service sending driverless cars onto public streets, that is a meaningful capability. It gives Tesla a hard stop for a misbehaving vehicle, a stolen one or one that surfaces a safety flag mid-shift, without depending on the car cooperating with a software-only "please pull over" instruction. It also concentrates a lot of authority in whoever holds the operator credentials.
Remote operations and safety-protocol code
Alongside the kill-switch, 26.4.5 adds new code for remote operations and safety protocols, plus refinements to the autonomous ride-hailing flow. The direction of travel is clear: the app is being built out as the control plane for unsupervised operation, including teleoperator handoffs, route management and rider-side safety interactions.
This lands after Tesla's teleoperation layer was publicly stress-tested. Earlier in May, unredacted NHTSA-published crash reports included two incidents Tesla attributed to its remote operators — a rare disclosure about how much human-in-the-loop scaffolding the service relies on. The sources don't draw a direct causal link between those crashes and 26.4.5, and neither do we — but the timing is worth noting.
Fleet growth: 9 to 39 in six weeks
CleanTechnica's mid-May tally puts the unsupervised robotaxi count at 39 vehicles, up sharply from earlier in the spring (Tesla is rolling out more unsupervised robotaxis):
| Date | Unsupervised robotaxis in operation |
|---|---|
| Early April 2026 | 9 |
| Early May 2026 | 26 |
| Mid-May 2026 | 39 |
The growth curve is real, but the absolute number is still small enough that a single bad day moves the safety statistics. Motley Fool's recent roadmap piece argues the meaningful scaling is still ahead, gated on the steering-wheel-less Cybercab variants spotted in production (Here's When Tesla's Robotaxi Rollout Will Really Ramp) — context to keep in mind, though it remains analyst speculation rather than a Tesla commitment.
EU angle: foundations before deployment
For European readers, none of 26.4.5 changes what's available in the EU today. Robotaxi remains US-only, confined to Austin, Houston and the Dallas area, and FSD (Supervised) in the EU is still gated by national and EU-level regulators. What 26.4.5 does is build out the operator-side control stack — remote disable, safety protocols, autonomous ride-hailing flow — that any EU autonomous deployment would need to demonstrate to regulators before approval. The kill-switch in particular is the kind of hard remote control European safety authorities have historically wanted to see baked in, not bolted on. It is foundation work, not a deployment signal.