Google has quietly introduced a new account setting called Search Services History. It is rolling out to users globally over the next few weeks, and for most people, it arrives switched on.
What it captures goes well beyond what Google used to save. The setting now stores the photos you take with Google Lens, your voice searches, recordings from Google Translate's speaking practice, live search conversations, and any files you upload during a search session. Google's official help page confirms this media is used to train its AI models, alongside your regular text search history.
The rollout follows your previous settings. If you had Google's older Web and App Activity tracking turned on, the new setting inherits that and turns on automatically. If you had it off, you stay off. But the default for most active Google users is on, which means the data collection starts without anyone clicking "agree."
There is a specific detail worth flagging. Once your media is used in AI training, Google disconnects it from your account. At that point, deleting your search history does nothing. The training data is kept for up to four years regardless. The window to stop this is before it happens.
To opt out: go to myactivity.google.com, look for the Search Services History tab, and switch it off. There is also a separate sub-option called Save Media that needs to be unchecked on its own. The tab may not appear in your account yet since the rollout is gradual, but checking weekly until it does is worth the two minutes.
This pattern, on by default with opt-out buried in settings, is now standard across the industry. X tried the same approach with its Grok AI in 2024 and faced regulatory pushback in Europe. Atlassian is doing something similar with its Jira and Confluence products starting later in 2026. The reasoning is always the same: better AI requires more data, and most users never change default settings.
For business operators, the relevant question is not just personal. If your staff uses Google Lens to photograph contracts, products, warehouse issues, or client sites during their working day, those images could be fed into Google's training pipeline. The same applies to any voice search done on a work device or account.
Google operates across Search, Maps, Translate, Lens, and News. That reach gives it a volume of real-world, multiformat data that most competitors cannot match. Turning it all into AI training material is a logical move for Google. The cost falls on users who do not know the setting exists.
European users may have additional protections under GDPR, which generally requires a stronger legal basis for processing personal data for AI training. Whether this specific setup satisfies that standard is an open question that regulators have not yet answered publicly. Users outside the EU are operating under weaker rules in most jurisdictions.
The practical advice is simple: check the setting now, turn off Save Media, and set a reminder to verify it again in a few weeks once the rollout completes for your account.