In the wake of Tesla’s first recorded autopilot crash, automakers are reassessing the risk involved with rushing semi-autonomous driving technology into the hands of distractible drivers. But another aspect of autopilot—its ability to hoover up huge amounts of mapping and “fleet learning” data—is also accelerating the auto industry’s rush to add new sensors to showroom-bound vehicles. The race to turn everyday vehicles into vast, data-collecting repositories is becoming a great technological opportunity for legacy carmakers. But in order to earn the enthusiasm of consumers, automakers and startups must first earn their trust.
Tesla’s much-discussed autopilot system consists of two elements: a suite of sensors and the software system that actually helps drive the car. In order to use the autopilot’s functions, a Tesla customer must pay for the software system. But even if they don’t choose this option, the built-in sensors are constantly recording information about its environment and how the driver is navigating through it. This enables every vehicle it makes to become a mobile data collector (whether the owner pays for the autopilot functions or not), and has allowed Tesla to begin catching up to Google, which has a significant head start on autonomous drive technology but a limited fleet with which to test and train its system.
Established automakers have taken notice of Tesla’s bold play and are seeking to leverage the same strategy across their even larger fleets. In the last year General Motors, Volkswagen, and Nissan-Renault have all signed partnerships with the sensor supplier Mobileye, and will integrate its camera-based “Road Experience Management” system into non-autonomous vehicles in order to jumpstart their crowdsourced mapping efforts. Toyota has announced an in-house effort to turn onboard cameras into mapping sensors, and the self-driving startup Comma.ai is even trying to encourage motorists to feed their driving data into its algorithms using a dashboard-mounted smartphone app.
Like Tesla, these companies lag behind Google in their autonomous technology development, but their massive scale could allow them to generate fleet and mapping data even faster than the Silicon Valley upstart. But extracting this personal data also runs the risk of upsetting the established relationship between carmakers and their customers. A 2015 Pew Research Center poll shows that Americans do not want cars to monitor their driving habits, even if such monitoring would lead to discounts on their auto-insurance rates.Of course certain technophile consumers will be eager to do whatever they can to help accelerate the advent of autonomous vehicles. But in order to foster this attitude, automakers and startups alike will have to be very careful about how they inform consumers about their data collection techniques and adhere to clear guidelines governing the use of such data.
So will they be up to the task? As with its rapid deployment of autopilot functions, Tesla’s fleet data collection is already demonstrating the risks involved as well as the rewards. In several instances, the electric car maker has publicly released characterizations of data captured by onboard sensors in order to fight the perception that its driver assistance system is unsafe.
This may surprise some users: Tesla’s Terms of Use (TOU) does not explicitly state that the company will release information captured by its vehicles when there is no clear legal need for it. That doesn’t mean they can’t, however, it all comes down to how you interpret the wording. Whether or not Tesla’s public disclosures of vehicle data fall under a reasonable interpretation of its TOU is a matter of legal interpretation. As a practical matter, however, the company’s failure to clearly disclose that it could publicly release characterizations of owner driving data could potentially lead to a backlash from privacy-minded consumers, especially as Tesla attempts to bring its technology to a mass market.
Others in the auto industry seem to be at least slightly more aware of the need for clear disclosure guidelines, as evidenced by the voluntary privacy principles drafted by the Auto Alliance trade group in 2014, which were signed by many of the largest automakers. But even these principles fall short of what consumer advocates have pushed for in the past following controversies involving vehicle data collection.
In 2011 one signatory to the Alliance principles, General Motors, generated a wave of controversy and calls for an FTC investigation when it quietly started tracking the vehicles of customers who had canceled their OnStar subscriptions. Following criticism, GM quickly reversed that decision, allowing customers who canceled OnStar service to opt-in to the tracking program rather than forcing them to opt-out. While the controversy has since subsided, GM’s new MobilEye camera-based mapping capabilities mean GM’s disclosure and opt-in policies deserve new scrutiny.
As of a 2014 revision of its policies, GM now includes five years of free OnStar Basic plan services “on most 2015 and newer vehicles,” meaning new vehicles using MobilEye technology could send mapping and driving data to GM’s servers for five years. OnStar plans must be activated, which technically makes even the basic service “opt-in.” But customers who verbally activate the OnStar system will be doing so without a full copy of the service’s privacy policy or TOU. Furthermore, using the OnStar RemoteLink smartphone app (which offers free services like remote door locking and unlocking) makes owners automatically subject to the OnStar privacy policy via the app’s End User License Agreement—even if they have not opted in to OnStar’s services.
Combining powerful data generation capabilities with poor disclosure and a weak opt-in system is a potentially risky recipe. The only situations where Auto Alliance signatories require “affirmative consent” for the collection and sharing of geolocation, biometric, or driver behavior information are when automakers wish to use that information for marketing or to share it for marketing purposes. Otherwise, vehicle data may be collected or shared without the owner approving a “clear, meaningful, and prominent notice” for a wide variety of uses, including “safety, operations, compliance, warranty, internal research, and product development.”
It doesn’t have to be this way, of course. Europe’s tough General Data Protection Regulation requires a far higher standard for transparency and customer choice. OnStar’s European customers are able to turn off vehicle tracking at any time with a single “privacy button” that does not affect the service’s ability to alert emergency responders during a crash.
GM doesn’t offer a “privacy button” on its US OnStar systems in part because it isn’t required to do so by law. Such lax legal standards are certainly part of the problem. But automakers that understand the long-term opportunities afforded by vehicle data should not risk souring such a powerful tool by sinking to the lowest legal standard; rather, they should be competing to offer consumers the most value and candor for their service as driving data generators.
This is a pivotal moment for the vehicle data economy. It’s up to consumers to demand stronger, clearer privacy standards—after all, automakers want our vehicle data more than we want any of the services they are offering in return. For the first time since the birth of the personal data economy, consumers may actually have the upper hand. Let’s not give it back to automakers just yet.