Tesla mocks Waymo over Philippines remote assistance
Robotaxi autonomy depends on teleoperation and exception handling, Regulators get driverless narratives with human backstops
Images
Tesla and Waymo are now openly litigating the robotaxi industry’s least marketable secret: “self-driving” is often a distributed system with a human in the loop—just not necessarily sitting in the car.
Business Insider reports that Tesla used California regulatory comments to take a swipe at Waymo after lawmakers criticized Waymo’s use of remote assistance workers in the Philippines. Tesla told the California Public Utilities Commission it employs only US-based staff to help its vehicles when they get stuck, arguing domestic operators are more familiar with local rules, have better “network connectivity,” and are less exposed to cybersecurity risks.
The jab follows a December 2025 San Francisco power outage that stranded some Waymo vehicles. Waymo said the blackout caused a spike in requests from its remote assistance (RA) system, overwhelming its capacity. Tesla contrasted that with its California ride-hailing service—currently using human drivers overseeing its Full Self-Driving software—which it said was not impacted.
Waymo, for its part, is trying to thread a regulatory needle: it wants the benefits of autonomy branding without accepting that remote assistance is, functionally, a form of driving. In a letter to Sen. Ed Markey, Waymo said RA agents do not directly drive vehicles but provide “advice and support” when the system encounters scenarios it can’t handle. The company told Markey it typically has around 70 agents on duty at any time, split roughly half in the US and half in the Philippines, with more complex cases handled by a specialist team in the US.
This is less a moral panic about offshore labor than a definitional fight about autonomy. If a vehicle can request help, receive guidance, and proceed, the operational reality is an exception-handling pipeline: detection, escalation, human interpretation, and a command channel back to the vehicle. The engineering constraints are mundane—latency budgets, communications redundancy, authentication, logging, and the safety case for what remote agents are allowed to do. The legal constraints are worse: who is the “driver” when a human provides steering-level intent but the vehicle actuates locally, and how does liability attach when the human is an employee of a contractor on another continent?
Politicians prefer the gray zone. A world with “no driver” is rhetorically convenient; a world with a driver you can regulate is administratively convenient. Remote assistance offers both: companies can claim driverless operations while regulators can demand processes, audits, and staffing ratios—creating a compliance perimeter around what is, at bottom, a labor model.
If California tightens robotaxi rules, the most likely outcome is not the death of remote assistance but its formalization: licensing categories for remote operators, prescriptive intervention protocols, and data retention requirements. The robotaxi race may be won not by whoever eliminates the human, but by whoever bureaucratizes him most convincingly.