Tesla launched its Robotaxi service in Miami on July 3, its first driverless ride-hailing market outside Texas and California, and it did something none of its earlier launches had: skip the safety-driver phase entirely. Rides in Miami are fully unsupervised from the first day of service, no human behind the wheel and no human in the front passenger seat either, a detail Tesla’s vice president of AI software, Ashok Elluswamy, confirmed directly on social media rather than leaving to inference from press coverage.
The initial service area covers roughly 10 to 14 square miles across western Miami-Dade County, deliberately excluding downtown Miami and Brickell, the densest and most pedestrian-heavy parts of the metro area. Access runs through the existing Tesla app to a limited but expanding pool of riders, with per-mile pricing designed to undercut Uber and Lyft’s current Miami rates rather than compete on service quality alone.
What makes the Miami launch technically distinct from Waymo’s approach, and from Tesla’s own earlier robotaxi markets, is the total absence of lidar or pre-mapped high-definition corridors. The service runs entirely on vision-based Full Self-Driving inference processed at the edge, cameras and neural networks making real-time driving decisions with no laser-based distance sensing and no fallback to a human operator if the system encounters something it can’t classify confidently. It’s the purest version yet of Tesla’s long-standing bet that cameras alone, without lidar, can eventually match or exceed the safety case built by rivals using multiple, more expensive sensor types.
That bet is being tested in a city built almost perfectly to expose its weakest point. In March 2026, the National Highway Traffic Safety Administration found that Tesla’s camera-only approach fails to detect and warn drivers appropriately under degraded visibility conditions such as glare and airborne obscurants, a finding that reads very differently once you consider where the company chose to launch its first fully unsupervised market next. Miami combines intense subtropical sun glare that bounces sharply off wet pavement, sudden and heavy thunderstorms that can reduce visibility to near zero within minutes, and standing water after nearly every afternoon rain, the exact conditions NHTSA’s own testing flagged as the system’s demonstrated weak spot.
Tesla is, in effect, running the most demanding real-world validation of its camera-only thesis in the market least forgiving of it, and doing so without the safety-driver transition period every prior Tesla robotaxi launch used to build a track record before removing human oversight. Whether that reflects genuine confidence in how far the system has improved, or simply a company racing to claim a fully-unsupervised-from-day-one milestone before a competitor does, is not yet answerable from the outside, and won’t be until the service has accumulated enough real-world miles in real Miami weather to say anything statistically meaningful.
It’s also worth sizing Tesla’s ambition against what it has actually built so far. Across all of its markets, Austin, the Bay Area, Dallas, Houston, and now Miami, spanning a combined geofenced area of roughly 1,190 square miles, Tesla is running approximately 34 vehicles in total, a fleet that has reportedly been shrinking rather than growing in recent months even as the company keeps adding new cities. Waymo, by contrast, operates roughly 3,000 fully driverless vehicles across more than ten US metro areas and delivers around 500,000 paid rides every week using a sensor suite that combines lidar, radar, and cameras rather than betting everything on vision alone. By that measure, Tesla’s Miami launch reads less as evidence it has caught up to the market leader and more as a geofence-expansion and headline-generation strategy, unsupervised-from-day-one status is a genuine technical claim, but a 34-vehicle fleet stretched across five cities is still a rounding error next to Waymo’s operating scale.
Manila and most other major Philippine cities share close to the exact profile NHTSA flagged as Tesla’s weak point: monsoon downpours that arrive with little warning, intense glare off flooded streets, and visibility conditions that already challenge experienced human drivers on a routine basis during wet season. That makes the Philippines a genuinely difficult market for camera-only autonomy of Tesla’s specific kind, even though chronic traffic congestion and strong ride-hailing demand make the underlying business case for autonomous taxis attractive here in theory.
The more immediately useful lesson for Philippine mobility founders isn’t about weather at all, it’s about regulatory risk-taking. Tesla’s decision to skip the safety-monitor phase entirely in a market its own regulator had just flagged concerns about is a live case study in how far a company is willing to push ahead of its regulator’s stated comfort level, and how a single serious incident under those conditions could reset the entire industry’s permitted pace of rollout, in the US and everywhere regulators watch US precedent closely. Any Philippine startup pitching investors on autonomous or advanced driver-assist features should study exactly how the LTFRB and other transport regulators react if Tesla’s Miami bet goes wrong, since that reaction, not Tesla’s own safety data, is what will actually shape the regulatory runway available to smaller players building similar technology.
Share this article