HomeAIGoogle Splits Gemini Chat History from AI Training

Google Splits Gemini Chat History from AI Training

Google is splitting the Gemini app’s chat History setting from a separate control over whether those chats are used to train generative AI models, so users can keep a history and opt out of training.

In a 16 November 2026 blog post, published at 17:00 GMT, Google said: “New settings let you keep your Gemini chat history, while opting out of generative AI model training.” Users will still be able to change how long history is saved, and to review or delete history manually at any time. “We’ll start rolling out these changes globally in the Gemini app in the coming weeks.” The post does not list a country-by-country timetable, and it does not set out a separate UK exception.

The Gemini Apps Privacy Hub, last updated 16 November 2026, describes the two controls in more detail. The History setting, when it is on, saves chats in a user’s History, including prompts, attachments, Gemini’s responses, tasks and tool calls, and, in some cases, data from connected apps. History is automatically deleted after 18 months, unless the user chooses a different period. A separate setting, also reached through Gemini History, “control[s] whether Google uses your chat content, including text & attachments, to train generative AI models.” If that training setting is on, Google “uses the content of your future chats to train generative AI models.” If it is off, “Your future chats won’t be used to train generative AI models, unless you provide feedback.”

The official wording is already about future chats. Ars Technica, in a 17 November article published at 16:50 GMT, goes further. “Importantly, this only applies to future interactions with Gemini,” Ryan Whitwam wrote. “If you have previously fed any data into Google’s AI, that’s already part of the training data.” Ars also wrote that reviewed chats already in the training data “won’t be retroactively deleted.”

The same Privacy Hub page says reviewed chats and related data “are retained for up to 3 years, even if you delete them or turn this setting off later.” Turning the training setting off, it adds, “doesn’t impact activity saved and managed by other Google services, including activity shared when you connect or use Gemini with those services.”

History off still leaves a two-month archive

Google is also changing what happens when History itself is switched off. The blog says that if the new History setting is off, “future chats will be saved for around two months; they’re used to respond to you, process your feedback, and protect Google, its users, and the public.” The Privacy Hub repeats that “around two months” figure, and says temporary chats, and chats held when History is off, are retained with the account for that period. Temporary chats, it says, “are not used to train Google’s AI models.” “If History is off and you don’t submit feedback, Google also does not use your future chats to train generative AI models.” Submitting feedback is different: Google then collects the feedback, the last 24 hours of chats, and any content in those chats, to maintain and improve Gemini Apps.

Ars Technica contrasted the new two-month window with the previous period, writing that if a user disabled history, Google “still retains that data for 72 hours (and 24 hours for the user)” today, and that “once the updated page is live, this duration will change to two months.” That 72-hour comparison is Ars’s. The official blog and Privacy Hub state the new “around two months” period, without restating the old 72-hour figure in the passages published with this update.

The change is a product-settings story, rather than a new industry-wide opt-out standard such as the shared AI-training opt-out that Google, OpenAI and Microsoft have backed. It is closer to other TSN accounts of Gemini controls, including a hidden “Full Access” setting tested in Gemini for Mac and Gemini’s use of Gmail and Calendar on Android Auto, a feature TSN noted was not offered in the UK. Google’s wording on this occasion is that the History and training split will start to roll out globally.

A split control is not a wipe of earlier training data

The useful change is real, and it is narrower than a slogan about “opting out of AI training” might suggest. Google is unbundling two jobs that used to sit on one switch: how long a chat is kept for the user, and whether the content of future chats is used to train generative models. People who want a searchable history can now, once the new settings appear, leave History on and turn the training control off. That is what the official blog promises.

It is not a promise to pull earlier chats out of models that have already been trained. The Privacy Hub’s own training FAQ is written in the future tense: when the setting is on, Google uses “the content of your future chats”; when it is off, “your future chats won’t be used,” unless the user sends feedback. Ars Technica’s reading, that material already used for training stays in the training data and that reviewed chats already there will not be deleted after the fact, sits on top of that official wording rather than contradicting it. The same page also says reviewed chats can be kept for up to three years even after a user deletes them or turns the setting off. Those limits belong in the same account as the new toggle.

The History-off change cuts the other way. Google is giving users a longer official reason to leave History off and still have a short archive, and it is also giving itself a longer window, of around two months rather than the 72 hours Ars describes as the current period, in which off-history chats remain available to respond, to process feedback, and for safety review. Temporary chats still sit outside training, on Google’s wording, if the user does not send feedback. The new settings are an improvement in control over future chats. They are not a receipt that earlier conversations have been forgotten.

Sources

Related stories

Google, OpenAI and Microsoft Back a Shared AI Training Opt-Out is the nearest TSN account of a broader training-opt-out effort, and is the next read.

Follow @tsnmediahub on X for breaking AI news as it happens.

Follow on X

Share this story

Latest stories

More in this category

Latest stories

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.