Google has introduced new privacy settings that enable the collection of users' media uploads for AI training, prompting concerns about data privacy and user control.
Recently, I received an email from Google with the subject line: 'New privacy settings for Search services.' This communication detailed a global update that is being implemented over the coming months, which will alter how Google manages users' Search history data.
Under these new settings, Google will retain various types of media, including images uploaded for reverse image searches and audio recordings made during interactions with Google Translate, to enhance its AI models.
Upon visiting the Search Services History page for the first time, I found that this new option was already activated. If users had previously turned off Google’s Web & App Activity and Search Personalization options, the feature would be disabled. However, the box allowing Google to save uploaded media for AI training was checked by default.
Once this feature becomes available on your account, you can opt out by going to Google’s My Activity page and selecting the Search Services History tab. This section reveals what Google records from your Search history and provides the option to disable the setting and delete your activity. To prevent your image uploads from being utilized for AI training, it is essential to uncheck the box labeled 'Save media.'
It is advisable to make this change promptly, as the implications of your media data being used for AI training are significant. A warning displayed when I disabled this feature stated, 'If your saved media is used to train our AI models, it is disconnected from your Google Account. This training data will be kept for up to 4 years, even if you delete the original activity.' This indicates a lengthy retention period for my uploaded images.
Davis Thompson, a Google spokesperson, commented via email that these new settings aim to help users receive more relevant results and revisit their searches, including visual and voice searches. He did not address WIRED’s inquiry regarding the default activation of this feature.
Google clarifies that it will be saving more than just text entries in Search. The description on the settings page states, 'Your saved media includes your images, files, and audio and video recordings from your interactions with Search services.' This encompasses various forms of content, such as images from Google Lens, recordings from Search Live or Translate speaking exercises, and voice search data.
AI models require diverse types of data, not limited to text, for improvement. By collecting a broader range of data from its extensive user base, Google may gain a competitive advantage in innovation over its rivals.
The company's vast user base across multiple services positions it uniquely in terms of data collection. Thorin Klosowski, a senior security and privacy activist at the Electronic Frontier Foundation, noted that Google's long-standing services have led users to become comfortable with the amount of data being collected, creating a sense of inertia that may prevent them from switching to alternative services.
The practice of requiring users to opt out of AI training is becoming standard across various platforms. Klosowski argued that an 'opt-in' approach would be a more responsible expectation from these companies, suggesting that users should actively choose to enable such features instead of having them automatically activated.
In an email sent to my testing account on June 23, Google framed the changes as providing users with 'even more control over saved history.' The email included examples of how saving media could be beneficial, such as allowing users to revisit past visual searches with Lens or continue conversations about songs heard during Search Live sessions. However, it is striking that Google did not provide comparable examples regarding the use of saved media for AI training.
This significant software change warrants careful consideration from everyday users. Ben Winters, director of AI and privacy at the Consumer Federation of America, remarked that it adds complexity for consumers who must now assess their comfort level with a tool they have been using for an extended period.
The burden of opting out of data training across numerous services can be overwhelming, leaving users feeling anxious about potentially overlooking important settings. Winters expressed concern that this trend may contribute to a growing sense of powerlessness and hopelessness among users regarding data protection, stating, 'There’s an increasing feeling of powerlessness and hopelessness about even trying to protect your data, because every little thing is going to be squeezed out of you.'
