Alphabet’s autonomous driving arm, Waymo, is preparing to use data collected from inside its driverless taxis — including video footage tied to individual rider identities — to help train generative AI models, according to a yet-to-be-released version of its privacy policy. The development, first spotted by independent researcher Jane Manchun Wong, has reignited long-standing questions about consent, surveillance, and the creeping reach of AI into everyday life.
Among the revelations is language confirming that Waymo may share user data to personalise advertising, a practice that, while not unprecedented, acquires new weight when coupled with in-vehicle cameras capturing the intimate and often unguarded behaviours of passengers.
The policy draft suggests that riders will be given the ability to opt out of having their data sold, shared, or used to train machine learning systems. However, it remains unclear how prominently this opt-out will be presented — or whether users will need to actively dig through menus and disclaimers to find it.
“Opt out of Waymo, or its affiliates, using your personal information (including interior camera data associated with your identity) for training [generative AI],” reads one portion of the leaked page.
Waymo confirmed that the feature is still in development, with spokesperson Julia Ilina stating that the privacy policy itself is not undergoing changes. Rather, the company will be introducing “a new opportunity” for users to opt out of AI training-related data usage.
Ilina sought to assure riders that personal data is used “to train AI models for safety,” such as ensuring cleanliness, aiding in emergencies, or locating lost property. “Any data Waymo collects will adhere to the Waymo One Privacy Policy,” she said, noting that the company does not share personal information with other Alphabet entities — which include Google and DeepMind — unless explicitly permitted by users.
Yet, in a digital climate of increasingly blurred lines between data collection, monetisation, and service improvement, the fact that Waymo’s AVs may soon be studying riders to improve algorithms raises concerns. As Wong pointedly noted, the inclusion of cameras linked to rider identities “ratchets up the creepiness factor”.
Waymo, the only autonomous vehicle operator in the U.S. currently offering paid robotaxi services, has grown rapidly in recent years. The company now logs more than 200,000 rides per week across Los Angeles, San Francisco, Phoenix, and Austin — a twenty-fold increase from just two years ago. Expansion plans for Atlanta, Miami, and Washington D.C. are in motion.
Despite this momentum, Waymo is not yet profitable — a fact that casts recent moves in a new light. In 2024, the company recorded an operating loss of $1.2 billion, filed under Alphabet’s vague “Other Bets” division. And while Alphabet has invested $5 billion into Waymo and raised a further $5.6 billion from outside investors, financial self-sufficiency remains elusive.
The push toward monetising in-vehicle data, particularly for AI training and advertising, may be a sign that the firm is seeking additional revenue streams beyond the farebox. Whether this proves a savvy strategy or a reputational misstep may depend less on technical capability and more on public perception and trust.
As Waymo cars roam city streets with increasing ubiquity, it remains to be seen how comfortable riders will be knowing their every move — and mumble — could become training fodder for the next generation of AI.
