So when does Gemini train on my data? I find Gemini's public statements on data use confusing: https://ai.google.dev/gemini-api/terms<p>"When you use Unpaid Services, including, for example, Google AI Studio and the unpaid quota on Gemini API, Google uses the content you submit to the Services and any generated responses to provide, improve, and develop Google products and services and machine learning technologies, including Google's enterprise features, products, and services, consistent with our Privacy Policy."<p>"When a Service is being offered for a fee, it is considered to be a paid Service (the "Paid Services"). When you activate a Cloud Billing account, all use of Gemini API and Google AI Studio is a "Paid Service" with respect to how Google Uses Your Data, even when using Services that are offered free of charge, such as Google AI Studio and unpaid quota of Gemini API."<p>This seems to make some sense, but then if I go to https://aistudio.google.com/plan_information I still see the following screen even after I add my payment info on Google Cloud: https://drive.google.com/file/d/1xqWTFTO_mjkvTbPfO971aKnlB7y2fVKN/view?usp=sharing<p>The screenshot shows for my account "DATA USED TO IMPROVE OUR PRODUCT: Yes" for the "Free of charge" panel on the left, which has a conspicuous star leading to a non-existent footnote. It is hard to tell if there is a catch somewhere. If Gemini ended up ingesting my data, I am not sure I would have much of a case against them given the explicitly ambiguous messaging on their end in my own account (might this be intentional?).<p>I like Gemini, but cannot commit to using them in production unless I have confidence they do not train on my data in AI Studio and through the API.<p>Can someone enlighten me?
There are multiple kinds of data.<p>1. The inputs and outputs to the model<p>2. How you click around<p>3. You click the feedback button<p>For (1), when paid, they will not use you I/o for improving the model. If you hook it to Workplace, they will not use it to train models, unless you ask them to build a custom, private model on you orgs files. There are risks to them reputationally and legally if they did and the model spit something back out to another customer.<p>The other two are pretty standard across all SaaS and normal expectations apply here