Google adds opt‑out toggle for Gemini training while keeping chat history
Google is rolling out separate toggles in Gemini that let you stop your chats from being used to train its AI models while preserving your conversation history.
Google announced a rollout of new privacy controls for its Gemini chatbot. The changes add separate toggles that let users keep their conversation history while opting out of having that data used to train the model. The update appears in the Gemini app and in the broader Google account settings.
Separate controls for training and history
According to Ars Technica, the new settings split the previous single history switch into two distinct options. One toggle continues to govern how long chat logs are retained for the user’s own reference. The second toggle, located under the Activity section of Gemini settings, lets users block Google from using their chats, attachments, and related activity to improve Gemini’s underlying models. The training toggle is an opt‑out, meaning it is on by default and must be turned off to stop data collection for model training. The history toggle still offers an adjustable limit, so users can decide how many past messages they want to keep visible.
Why the change matters to small businesses and nonprofits
Many small organizations rely on Google Workspace tools, including Gemini, for day‑to‑day communication and quick content creation. Before this update, turning off history was the only way to prevent data from entering the training pipeline, but that also erased the ability to resume a conversation later. Losing chat continuity can be disruptive for teams that reference prior discussions, draft proposals, or store uploaded files within the chat.
The new opt‑out gives these groups a way to protect sensitive information, such as client details, grant proposals, or internal policies, while still benefiting from Gemini’s assistance. However, the article notes that any data shared with Gemini before the toggle is activated may already be part of the model’s training set, and Google can retain that data for up to three years. Organizations should treat that as a reminder to review what they have already fed into the system and consider whether older content needs to be removed or anonymized elsewhere.
How to enable the opt‑out toggle
To stop Gemini from using your chats for model training, follow these steps:
- Open the Gemini app on your device.
- Tap the settings icon and choose Activity.
- Locate the toggle labeled Model training (or similar) and switch it off.
- If you prefer to manage the setting from your Google account, go to Data and Privacy > My Activity > Other activity > Gemini Apps Activity and turn off model training there.
- Adjust the Chat history toggle separately if you want to keep a limited log of past messages.
These steps can be repeated for each account that accesses Gemini, so team administrators should verify that the setting is applied to all shared or service accounts used for business purposes.
Best practices for ongoing data privacy
Even with the new toggle, good habits are still important. Here are a few recommendations for organizations that use Gemini or any generative AI service:
- Review the content you share in chats and remove any unnecessary personal or confidential details.
- Periodically audit the My Activity page to confirm that the training opt‑out remains active.
- Keep a record of when the toggle was changed, especially if you need to demonstrate compliance with internal policies or external regulations.
- Consider using separate accounts for public‑facing queries and internal work to limit the exposure of sensitive data.
- If you are uncertain how AI fits into your workflow, see our plain‑English AI roadmap for guidance.
By combining the new opt‑out feature with disciplined data handling, small teams can continue to leverage Gemini’s capabilities without sacrificing control over their proprietary information.
Sources
This post was drafted with AI from the reporting linked above and published by The Jones Foundry. For full details, read the original sources.