When you upload any media to Google's search services, it's now being used to train artificial intelligence models. The result? Less privacy — unless you opt out.

Google quietly updated its search privacy settings last month, as announced in an email, and automatically opted all its customers into its expanded AI training. The updated documentation states: "You can choose whether Google saves media to your Search Services History when you're signed in. Saved media includes your images, files, audio, and video from your interactions with Search services. Your media may be used to improve your experience on Google services, like letting you revisit your past visual searches. Saved media may be used to develop and improve Google's AI models and technologies, as well as the Google services that use them."

This means any media generated by your search using Google's services — including screenshots taken with Circle to Search, photos you've taken, voice searches via Search Live or Google Translate — can now be saved and used for training large language models (LLMs) like Gemini. Previously, Google mainly used text-based search queries and other interactions for AI training, but the expansion to visual and audio data marks a significant escalation in data collection.

What does this mean for your privacy?

This new language essentially grants Google permission to use your personal media for any purpose, including future searches and training AI. When you combine that with the company's existing targeted advertising systems, a picture emerges of steadily diminishing privacy. For example, if you discuss leaving your job with a Google chatbot, that conversation could be used to train the LLM. If you share specific details—like your company name, your boss's identity, and reasons for dissatisfaction—the AI learns from that information. Later, your boss might ask the same AI about employee loyalty, and the response could inadvertently expose your plans.

Voice data is especially sensitive. Your voice recordings could be used to train voice cloning models, potentially enabling malicious actors to create fake audio clips of you saying things you never said. While Google has safeguards, the mere act of storing such data creates new attack surfaces.

How to opt out and protect your data

Fortunately, you have some control. The most important step is to disable the feature that saves media for AI training. This can be done through two main pages in your Google account settings: Search Services History and Search Services Personalization. Here's a step-by-step guide:

1. Open the Google app on your mobile device (or go to myactivity.google.com on desktop).
2. Tap your profile image in the top right corner and select "Search history".
3. On the Search History page, make sure "Save media" is unchecked. Then tap the top entry for Search Services History. If it shows as "On", tap the button and turn the service off.
4. Next, go back, tap your profile image again, and select "Search personalization". Tap the top entry and turn off the service.
5. Finally, visit myadcenter.google.com and ensure "Personalized Ads" is disabled.

If your device doesn't allow you to disable these from the app, use a desktop browser: go to myactivity.google.com, click Search Services History, and uncheck the box for Save media. You can also configure automatic deletion of saved data after 3, 18, or 36 months. However, the safest option is to completely disable media saving.

The bigger picture: data as fuel for AI

Google's move is part of a broader industry trend where tech giants are scraping ever more personal data to improve their AI models. The company's rationale is that more data leads to better, more helpful AI assistants. However, critics argue that this comes at the cost of user consent and privacy. While Google claims it does not sell your data, training AI models effectively creates new digital representations of your behavior, which can be exploited in unforeseen ways.

The expansion to include images, videos, and voice adds a new layer of risk. Consider the scenario where you upload a photo of your home to search for similar products. That image could become part of the AI training set, potentially allowing the model to recognize your home in other contexts. Voice samples, even anonymized, can be used to improve speech recognition but also carry the risk of re-identification if linked with other data.

For those who value privacy, the solution is clear: opt out. The steps described are straightforward and should take only a few minutes. Additionally, consider using locally installed AI models for sensitive tasks, as they keep your data entirely on your device. While Google's services offer convenience, the trade-off in privacy is becoming steeper with every policy update.

As AI becomes more embedded in daily life, staying informed about how your data is used is essential. This latest change from Google is a reminder that you should regularly review your account settings and make conscious choices about what you share.

Source:ZDNET News