GOOGLE IS NOW TRAINING ITS AI ON YOUR UPLOADS, YOU'RE OPTED IN ALREADY
Images, audio, and video uploaded to search tools like Google Lens can be used for training unless you uncheck a box.
by editor4 min readcomments soon
Google quietly changed the rules. Every photo you snap through Google Lens, every file you upload to its search tools, every audio and video clip you hand over can now be used to train the company's AI models. And you're automatically opted in.
The media types covered include "images, files and audio and video recordings", according to the policy update. That covers Google Lens, where users regularly upload photos to get information about objects, landmarks, or text. It also applies to other search-tool uploads, and a voice recording could now feed the training pipeline. The one notable carveout is personal Google Photos. Your private albums and backups are not included, which suggests the change is limited to uploads made in the context of a search task rather than storage.
THE DEFAULT IS OPT-IN
Every user is automatically opted into this arrangement. Google says you can opt out by visiting the Search Services History page and unchecking the box. That's the entire process. Find the page, find the checkbox, uncheck it.
The opt-out design matters because the typical user has no idea this setting exists. The Search Services History page is not a place most people visit regularly; it is a privacy dashboard that surfaces data Google has collected, not a control panel for training preferences. Hiding the toggle there while making the policy opt-in by default is a deliberate friction point. It means Google gets to point to an opt-out mechanism if regulators ask, while most users will never find it.
WHAT THIS SAYS ABOUT AI TRAINING NORMS
The claim is not a direct quote from Google. It is an observation from the reporting. But it captures the prevailing attitude in the industry. Training data is the lifeblood of large models, and the companies building them have been aggressive about sourcing it from anywhere they can legally reach. Google's move here is consistent with that playbook: expand the pipeline by reclassifying user uploads from ephemeral inputs to training material, then make the exit hard to find.
What makes this notable is the specificity of the data. Uploads to Lens and similar tools are often highly contextual. Those are not the kind of generic public data AI companies typically scrape from the web. They are personal, situational, and sometimes sensitive. Google is now folding them into the same pool used to train models that will be deployed back at the public.
The fact set includes a workaround: you can turn off AI overview results by adding "-AI" before a query. It is a hack, not a setting. You have to remember to type the minus sign and the letters into every single search bar.
WHAT YOU SHOULD DO RIGHT NOW
If you want to keep your uploads out of Google's training pipeline, the only path is the Search Services History page. Log into your Google account, go to the search history section, find the checkbox, and uncheck it. That stops future uploads from being saved for training.
The tip about adding "-AI" before a query does work, for now. It strips AI overviews from that specific search. If you want to avoid AI-generated answers in results, you need to add those three characters to every query you run.
And because the opt-in is automatic, the company can claim it has a legal basis for the use, even if the ethical basis is thinner. The gap between what users think they are doing and what Google thinks they are doing is exactly the kind of mismatch regulators are beginning to probe.
For now, the practical advice is straightforward. If you use Google Lens or any search tool that accepts media uploads, go find that checkbox. If you are tired of AI overviews, get comfortable with "-AI". The tools are there, but Google is not going to hand them to you.
what did you make of it?
more from ai
ai
TSMC ADDS $100 BILLION TO ARIZONA CHIP BET, TOTAL HITS $265 BILLION
The additional investment will build at least four more 2nm fabs and advanced packaging, bringing the company's total US commitment to $265 billion.
ai
META WILL ALERT PARENTS IF TEENS DISCUSS SUICIDE WITH META AI
The opt-in feature flags self-harm references in chatbot conversations, with human review before any notification is sent.
ai
ROBLOX'S "BUILD" LETS ANYONE MAKE A GAME FROM THEIR PHONE WITH AI
The new toolset, launching July 28, turns text prompts into playable experiences and puts game creation on iPhone and iPad.
ai
ZOOX REALLS ENTURE ROBOTAXI FLEET OVER SMOKE DETECTION FAILURE
A robotaxi drove into an active fire scene obscured by smoke. NHTSA called emergency scenes not edge cases.