The settlement at a glance
The $220 million figure is the headline, but the structural remedies may matter more. Under the agreement, a third party will be required to audit GM’s privacy programmes to ensure the practice does not recur. That independent oversight is unusual in data-privacy settlements, which often rely on corporate promises rather than external scrutiny.
The case was coordinated across 36 states, a sign of how seriously regulators viewed the allegations. Washington state Attorney General Nick Brown led the effort, and his office has been explicit about the mechanics of what it says went wrong.
How OnStar Smart Driver worked
Millions of drivers of GM vehicles in the US were enrolled in OnStar Smart Driver between 2016 and 2024. The pitch was straightforward: the system would help you improve your driving, and your data would not be sold. That second promise is at the heart of the case.
In reality, the system collected precise geolocation and driving-behaviour data and sold it to data brokers including LexisNexis and Verisk. The data included hard braking events, speeding and late-night driving. These are exactly the metrics an insurer would want when assessing risk, and that is precisely how they were used.
GM terminated its deals with LexisNexis and Verisk two weeks after the original New York Times story was published. By then, years of driving data had already changed hands.
The enrolment allegations
The allegations concern how drivers ended up in the programme. Brown alleges that many drivers were automatically signed up for Smart Driver when they subscribed to OnStar services. In some cases, dealership employees accepted terms and conditions on customers’ behalf without their consent.
GM also heavily incentivised dealerships for OnStar sign-ups, with employees receiving bonuses based on how many customers they enrolled. When a bonus depends on enrolment numbers, the pressure to sign people up quietly rather than explain the terms is obvious. The allegation is not that dealers were rogue operators, but that the incentive structure rewarded volume over consent.
The 370,000-driver problem
GM’s OnStar app allegedly told 370,000 drivers they were not enrolled in Smart Driver when they were, and their data was being sold. When GM unenrolled those drivers without notice, brokers and insurers retained the sensitive driving data.
Drivers who checked their status were given incorrect information. When the error was corrected internally, the data did not disappear from the brokers’ systems. The people affected had no way of knowing their driving history was still circulating in insurance pricing models.
What insurers did with it
The data was used to compile credit reports on drivers. Some insurers used those reports to set rates, and in some cases to deny insurance altogether. A hard-braking event on a wet Tuesday morning could follow a driver for years, long after the incident itself was forgotten.
Insurance pricing is already opaque to most drivers. Adding a hidden data stream that drivers did not knowingly supply makes it harder still to challenge a quote or understand why one insurer is so much more expensive than another.
Regulators step in
The settlement is not the only regulatory action. In January 2025 the Federal Trade Commission banned GM from sharing geolocation and driver behaviour data with consumer reporting agencies for five years. The FTC also required GM to obtain express consent before collecting connected vehicle data and to let owners disable collection of precise geolocation and driving behaviour data.
Express consent means an active, informed yes rather than a buried clause in a subscription agreement. The ability to disable collection is equally significant, because it gives drivers a genuine off switch rather than a promise that data will be handled carefully.
GM’s response
GM has defended connected data as a feature rather than a liability. A spokesperson told Reuters: “Safer, more helpful and personalized experiences are enabled by connected vehicle data, and GM is committed to being clear about what information is used, how it’s protected, and the choices customers have about what they share.”
The spokesperson said the automaker has strengthened privacy with the system. That is the standard corporate response after a settlement, and it is not unreasonable on its face. Connected data genuinely does underpin features drivers value, from crash notification to remote diagnostics. The question is whether the consent architecture around that data is designed to inform drivers or to move them past a screen as quickly as possible.
GM’s ambitions in connected and assisted driving have not slowed. The company has set out plans for eyes-off highway driving due in 2028, a programme that will depend on exactly the kind of location and behaviour data at issue here, as detailed in GM’s Eyes-Off Highway Driving Coming in 2028: Inside the Plan to Skip Level 3. It has also been making the case for its electric truck strategy, covered in GM Says Electric Trucks Have a ‘Huge Application’ as It Defends EV Pickup Future. Both directions increase the volume of data vehicles generate.
What this means for connected cars
The GM case is unlikely to be the last of its kind. Modern cars are effectively smartphones on wheels, generating continuous streams of location, speed, braking and camera data. That data has real value to insurers, marketers and third parties, and the commercial temptation to monetise it is considerable.
The settlement also shows that the gap between what a company says about data and what it does with it can persist for years before anyone notices. The original New York Times investigation did the noticing. It took a 36-state coalition, a federal ban and $220 million to produce a remedy. Drivers who want to know how their own car treats their data would do well not to wait for the next investigation. The practical starting point is the privacy menu in the car itself: check which connected services are switched on, read what each data-sharing option actually covers, and remember that a free trial can quietly become a data pipeline.





