A Waymo robotaxi has careered into a creek in Phoenix, Arizona, prompting a recall of thousands of vehicles and triggering a UK safety review as regulators scramble to understand the failure. The incident, which occurred late Tuesday local time, saw the self-driving Jaguar I-Pace veer off a residential road and plunge into a shallow waterway. No injuries were reported, but the spectacle of an empty car slowly filling with murky water has become a potent symbol of the perils of autonomous mobility.
Waymo immediately recalled 672 vehicles, primarily in the Phoenix and San Francisco fleets, and paused operations in affected areas. This is a significant setback for the Alphabet subsidiary, which had been touting its safety record after years of testing. The UK’s Centre for Connected and Autonomous Vehicles (CCAV) has launched a rapid review of all robotaxi operations in the country, particularly focusing on mapping failures and sensor redundancy.
The question on everyone’s lips: can we trust these machines with our lives? Let me break down what we know so far. The vehicle was operating without a safety driver, a common practice for Waymo since 2020 when they removed human overseers from most Phoenix routes.
According to preliminary telemetry data, the car’s perception algorithms identified a low-lying obstacle ahead—possibly a dumped tyre or a stray animal—and initiated an aggressive avoidance manoeuvre. The path planning system then failed to recognise the creek as a non-navigable area, likely because the onboard high-definition map was outdated or the lidar sensors misread the water’s surface. The car entered the water at roughly 8 mph, a decision that no human driver would have made.
Waymo’s engineering team has since identified a software bug in the hazard detection module. The recall will involve a physical inspection and a software update for all affected vehicles, a process that could take weeks. For the UK, this incident hits at a delicate time.
The government has been actively courting autonomous vehicle companies, offering regulatory sandboxes and tax incentives. London’s streets are already home to test fleets from Waymo rival Cruise, as well as British startups like Oxa and Wayve. The CCAV review will likely demand that all UK-based robotaxis operate with remote monitoring until the root cause is proven.
But the deeper issue here is the user experience of society: we are now beta testers for technologies that can kill. The narrative of inevitable progress is crashing against the reality of edge cases. A creek in Arizona may not sound like a typical urban hazard, but that is the point.
The physical world is messy. Construction zones, unmarked roads, and weather conditions all confound the most carefully trained models. The Phoenix incident also raises questions about public trust.
A 2023 survey by the UK’s Transport Research Laboratory found that only 43% of Britons would feel safe in a driverless car. After this, that number will likely drop. We are not merely adopting a technology; we are adapting our entire built environment to accommodate it.
The cost of that adaptation—both financial and psychological—must be accounted for. Waymo has handled the recall commendably by being transparent, but they have not yet explained why the sensor fusion stack failed to classify water as an obstacle. Water has unique reflective properties that can blind lidar.
The solution might involve a costlier suite of sensors like thermal imaging or sonar. But that adds expense and complexity, potentially slowing deployment. The UK review should demand radical transparency.
The government must mandate that all autonomous vehicles publish their safety data, including near-miss events, to a central regulator. It is not enough for companies to self-certify. The public deserve to see the code that decides whether to swerve or brake.
We have seen what happens when tech firms operate in secrecy: it eventually leads to tragedy. This incident is not a failure of technology per se, but of risk management. The promise of self-driving cars is that they will reduce the 1.
3 million road deaths annually worldwide. But if we rush to deployment without rigorous oversight, we risk trading one set of errors for another. The robotaxi that took a swim in Phoenix should serve as a wake-up call.
We must proceed with caution, not hubris. The algorithm is watching you. But who is watching the algorithm?
