In the continuing saga of consumer data being collected and used in ways people might not have anticipated, AI technologies provide a unique source of this information. The discussions and conversations you have with them, the questions you ask, the information you share, it’s all ingested. Google, for example, uses your regular search data to train the AI. Platforms like ChatGPT also regularly scrape user data. Additionally, a Stanford study identified six US companies offering AI services that not only train systems with the collected data, but also retain the information indefinitely.
The study suggests that when you use something like Google Gemini, all the information it learns and stores feeds into something more, likely far beyond personalized ads. In Google’s terms, it uses the data to provide and maintain services, develop new ones, personalize them, communicate with customers, measure performance and protect Google, users and the public. If targeted advertising seems the most alarming, it’s because you definitely feel like your phone, or someone on the other end, is listening to you.
However, they are so accurate because of the wealth of private information companies have collected about you. The data comes from a combination of companies and data brokers, companies that profit from the aggregation, disclosure and sale of personal information. It is collected through the digital services you use, like AI tools or virtual assistants, Google Gemini being a great example. And as Google explains, all of this is used for more than just targeted ads.
Google says Gemini data is not used for advertising or training purposes, but that may not be true
Primakov/Shutterstock
Google says your Gemini app chats “are not used to show you ads” and maintains it doesn’t train AI with your data. It is always healthy to approach these claims with a certain degree of skepticism. Not because the companies involved are outright lying to you, although they certainly do that sometimes, but because the reality is not black and white. Google’s claims are only half true.
This simple assertion can be questioned by simply looking at the myriad ways Google uses your data, and that’s before considering its publicly announced plans to run ads on Gemini (via Adweek). Looking at the facts, Google’s AI models can be trained on Gemini inputs and outputs, which can include email snippets or context, Google Workspace files, and other Google product data. Gemini app chats may not be used in this way, but other data is, including emails you send through the service.
When it comes to serving ads, Google explicitly states that when you interact with a voice assistant, it “may use the text from those interactions to inform about your interests for ad personalization.” However, this is Google Assistant, not Gemini directly. But let’s say for a moment that Google’s claims are true and that they are 100% true, unequivocally not using any of the data these services collect to train AI, serve targeted ads, or do anything shady. The amount of data collected just from Gemini is concerning, especially since it now supports Android phones and mobile devices by default.
What information does Gemini collect?
Nwz/Shutterstock
Google explains what Gemini collects in detail: Using Gemini apps stores information about the content generated by the tool, details about the steps or tasks needed to perform actions for you or respond, information from your apps, browsers and devices, additional information, location information and subscription information. There’s also the information “you provide” to Gemini, such as what you say to it, what you share, transcripts and recordings of interactions, including video, audio, screens you share and much more. It’s an uncomfortable truth about using Google Gemini, but it comes with huge privacy tradeoffs.
Google also says in the section about how it uses all this data that human reviewers can preview some of this information. It even advises against entering confidential information or anything sensitive that “you wouldn’t want a reviewer to see” or anything you wouldn’t want “Google to use to improve (its) services, including machine learning technologies.” It is also unclear how long some of the collected data is retained. Google shares timelines, but this varies depending on the nature of the data and how it is used.
Chats are stored for up to 72 hours, while app data can be stored for up to 18 months. You can choose to automatically delete Gemini app activity at different intervals or manually delete Gemini app chats “at any time,” but Google says some data is retained “longer if necessary” or, in some cases, until you delete your Google account. Outside of Gemini, you can also view or delete data from the basic (classic) version of Google Assistant.
