Uber to turn its millions of drivers into a global sensor network to fuel the self-driving industry

Uber is preparing to flip the switch on its most ambitious infrastructure project yet: transforming its global fleet of millions of human drivers into a massive, roaming sensor grid. The strategy, revealed by Uber CTO Praveen Neppalli Naga at the StrictlyVC event on April 30, 2026, aims to solve the "data bottleneck" currently stalling the autonomous vehicle (AV) industry.

Editorial credit: gguy / Shutterstock

By equipping everyday ride-share vehicles with specialized sensor kits, Uber intends to become the "data layer" for the entire self-driving ecosystem.

From ride-hailing to "AV Cloud"

Uber has already transitioned from building its own self-driving cars to becoming a marketplace for others. Now, the company is moving deeper into the stack with Uber Autonomous Solutions and the newly launched AV Labs.

The long-term vision is to create an "AV Cloud" - a searchable library of labeled, real-world sensor data. Instead of AV startups spending billions to deploy their own test fleets, they can simply "query" Uber’s network for specific scenarios, such as "a school crossing in San Francisco during a rainstorm at 3:00 PM."

The "shadow mode" advantage

Beyond just selling data, Uber is offering a platform for partners to test their AI in "shadow mode." This allows an autonomous model to run virtually during a real Uber trip, comparing how the AI would have steered against the actions of the human driver, all without putting an actual robotaxi on the road.

"The bottleneck is no longer technology; it’s the capital required to collect data," Naga explained. "We want to democratize this data to help our 25+ partners, including Wayve and Waymo, scale toward Level 4 autonomy faster."

Internal AI revolution

The shift toward being a data-first company is also being driven by a radical internal overhaul. Uber revealed that 95% of its engineers now use AI coding tools like Claude Code, and approximately 70% of all committed code at the company is now AI-assisted.

In a surprising admission, Naga noted that Uber is already seeing 1,800 code changes per week written entirely by autonomous AI agents, a transition he calls "agentic software engineering."

Privacy and regulatory hurdles

The plan to turn private contractor vehicles into "rolling data machines" isn't without its challenges. Uber is currently navigating a maze of state-by-state regulations regarding mobile sensors and has launched a Road Data Collection Privacy Hub. The company has committed to automatically blurring faces and license plates before any data enters the AV Cloud.

For the millions of Uber drivers, the shift raises questions about compensation and the future of their roles. While Uber insists the goal is to "democratize" data rather than monetize the drivers' every move, the reality is clear: the very people currently driving for Uber are now being used to train the machines that will eventually replace them.