In short
Google is updating Gemini so users can separate chat history from AI training consent, while keeping temporary and history-off chats for about two months instead of 72 hours. The change makes Gemini’s controls more like ChatGPT and Claude, but it also means “off” no longer means immediate deletion.
- Gemini will split chat history from AI training consent.
- Temporary and history-off chats will be kept for about two months.
- Google says the longer retention is for safety and security.
- The change is rolling out globally over the coming weeks.
- Competitors like ChatGPT and Claude already separate history from training.
Google is changing Gemini so users can separate chat history from AI training consent, while also extending retention for chats with history turned off from 72 hours to about two months. The shift, which TechCrunch said is rolling out globally over the coming weeks, makes Gemini’s privacy controls closer to rivals such as ChatGPT and Claude — but it also means even the strictest setting will keep conversations much longer than before.
What Google is changing in Gemini
The core update is a structural one: Gemini’s chat history setting will no longer be bundled with the decision to let Google use conversations for model training. That matters because the older Keep Activity setup tied those two choices together, forcing users into a single tradeoff instead of two separate ones.
Under the new system, users can keep chat history enabled without allowing their chats to be used for training, or they can turn history off while still withholding training consent. According to TechCrunch, the change is part of a broader update to Gemini’s privacy controls, and it is being pushed out worldwide over the coming weeks.
Google has also extended how long temporary conversations and history-off chats are stored. TechCrunch reported the new retention window at about two months, up from 72 hours. Crypto Briefing said the default for history-off chats is now around two months for service feedback and safety, while Notebookcheck pointed to Google’s updated Gemini Apps Privacy Hub, posted on Nov. 16, 2026, as the source of the new policy language.
How the new controls work
The practical effect is simple: users now get a cleaner split between privacy and product improvement. But the details still matter, because the new setup does not mean Google stops storing chats altogether.
Google says the longer retention is meant to help with safety and security as AI systems become more capable. In the company’s view, preserving more recent exchanges can help with abuse prevention, service reliability, and security review. TechCrunch also reported that Google cited possible system failures as one reason to keep chats longer.
Google’s updated privacy page says chats may be used to respond to users, process feedback, and protect services and users. Notebookcheck went further, reporting that the hub now says Gemini data may be retained “around two months,” and may be used “to respond to you and to protect itself, its users and the public.”
How is this different from the old Keep Activity setting?
It is different because Keep Activity no longer acts as a single master switch for both history and training. Before this change, turning history on or off also determined whether Gemini conversations could be used for training. That bundle made the user choice blunt: preserve chat logs and allow model use, or avoid both.
The new model decouples those decisions. That makes Gemini more flexible, but it also makes the product easier to compare against competitors that already separate history from training permission.
Why it matters
This is about more than a menu change. Gemini sits inside a broader consumer privacy debate that has become central to AI adoption: users want the convenience of chat history, but many do not want their conversations fed back into model training. Google is now acknowledging that those are not the same decision.
For readers, the significance is twofold. First, Gemini becomes easier to manage for people who want to keep a useful record of their chats without widening the data used to improve the model. Second, the company’s decision to retain history-off chats for roughly two months raises the bar for what “off” means in practice. The setting is no longer a near-immediate purge.
For the market, the update shows how AI companies are converging on a common privacy pattern. Crypto Briefing noted that ChatGPT and Claude already separate history from training, so Google’s move narrows a feature gap that has been easy for competitors to highlight. The remaining difference is in the details of retention and feedback use, where each company still draws its own line.
For the region and regulators, the change may invite closer questions about data minimization, storage duration, and how retention policies interact with emerging AI oversight. Even if Google is only aligning with industry practice, longer default retention could still matter if privacy rules or user expectations tighten.
What exactly stays on, and for how long?
The new policy is best understood as a set of buckets rather than a single on/off switch. Different kinds of Gemini chats now fall into different retention periods depending on whether history is enabled and whether the user sends feedback.
| Setting or action | Retention period | What changes |
|---|---|---|
| Temporary chats / history off | About two months | Up from 72 hours, per TechCrunch |
| Chat history on | 18 months by default | Crypto Briefing says the period is user-adjustable |
| Temporary ChatGPT chats | 30 days | Benchmark cited by TechCrunch |
| Incognito Claude chats | 30 days | Benchmark cited by TechCrunch |
That table shows the most important comparison in the story. Google is not alone in keeping some conversations for operational reasons, but its new baseline for history-off chats is materially longer than the old Gemini setup. At the same time, history-on chats are being kept for much longer than temporary chats, which reinforces the idea that Google sees those logs as a durable product feature, not just a short-term cache.
What happens if a user turns history off?
Turning history off does not mean the conversation disappears immediately. According to TechCrunch, those chats will now be retained for around two months rather than 72 hours. Google says temporary conversations are not used to train AI models, but they can still be kept for safety, security, and service-related purposes.
Crypto Briefing added an important caveat: even when training is off, a feedback interaction can still support model improvement. In other words, the user’s explicit choice not to train the model is respected in general, but a thumbs-up, thumbs-down, or other response rating can still carry separate data implications.
Background: why AI chat apps keep logs at all
AI chat systems routinely store some conversation data for a few basic reasons: to improve responses, detect abuse, troubleshoot failures, and review safety issues. Those purposes are now standard across the sector, even if companies describe them in different language and use different retention windows.
Google’s Gemini update sits inside that broader pattern. The important shift is not that retention exists, but that Google is making the privacy logic more explicit. The company is also trying to keep pace with rivals that already separate user history from model training permission.
Crypto Briefing reported that ChatGPT and Claude already offer that distinction. That means Google was under pressure to remove a clunky privacy design that made Gemini feel less transparent than competing products. The new settings do not eliminate the retention debate, but they do make the choice easier to understand.
There is also a chronological wrinkle. TechCrunch said an archived Gemini Privacy Hub from Nov. 12, 2026 still showed 72-hour retention, while an earlier version dated Oct. 26, 2026 also pointed to the older policy. Notebookcheck said Google updated the Gemini Apps Privacy Hub on Nov. 16, 2026. Taken together, the reporting suggests the change was applied recently and then began surfacing publicly only after the privacy hub update.
What changes for users in practice?
The biggest difference for most people is that the privacy setting is now easier to reason about, but the retention policy is longer than before. Users who care about conversation history and model training can now choose those preferences independently, rather than treating them as one combined decision.
That matters for three common user groups:
- People who want convenience: They can keep history on for easier reference without automatically opting into training.
- People who want tighter privacy: They can turn history off, but they should understand Google still retains some chats for about two months.
- People who use feedback tools: They should assume ratings or feedback may have separate data consequences even when training is disabled.
Crypto Briefing said adult personal accounts will have chat history on by default, while training will be off by default. It also said history-on chats are retained for 18 months by default, though users can adjust that period. If that report holds across the rollout, the default experience will be designed to preserve utility while limiting model-training permission.
Users who had Keep Activity off will see both of the new settings disabled by default, according to TechCrunch. Those who had Keep Activity on will keep chat history enabled by default. That suggests Google is trying to preserve existing user intent while shifting the interface underneath it.
Not every report describes the same details
The broad story is consistent across outlets, but some of the finer points differ. TechCrunch framed the new storage period for temporary and history-off chats as about two months, and compared that with 72 hours under the old system. Crypto Briefing emphasized the default state for adult personal accounts and said history-on chats are kept for 18 months. Notebookcheck highlighted the updated privacy hub language and reported a potential three-year ceiling for chats reviewed by service providers.
Those differences are not necessarily contradictions. They may reflect different parts of the same policy: default retention, adjustable history retention, and separate handling for chats that are manually reviewed by service providers. But they do show why users should read the policy carefully rather than assuming one number covers every case.
What did Notebookcheck add to the picture?
Notebookcheck’s report added two important details. First, it said that if a user leaves a response rating or sends feedback on a response while History is off, Google may also collect the last 24 hours of chats, including attachments. Second, it said service providers may keep reviewed chats for up to three years, separated from the user’s account.
If accurate, that means some chats can live much longer than the headline retention figure suggests. The policy would then have multiple layers: a short-to-medium storage window for normal conversations, a longer default for history-on chats, and an even longer reserve for reviewed material handled by service providers.
Why Google may be doing this now
Google has not publicly explained why the change is happening now beyond the safety and security rationale. Still, the timing points to a few plausible pressures that are visible from the reporting.
One is competitive parity. Once ChatGPT and Claude offered a clean split between history and training, Gemini’s bundled setting looked dated. Another is user trust. AI companies have learned that privacy controls are not only a compliance issue but also a product issue, and settings that feel confusing can become a reputational problem.
A third is operational. As Google said, AI systems are becoming more capable, and that can create new safety and security concerns. Longer retention can help with investigating failures, suspicious behavior, or moderation issues. The challenge is that those benefits come with a larger privacy footprint, especially when users may assume that “history off” means a near-erasure policy.
What happens next
The rollout is expected to continue globally over the coming weeks, so the final user experience may vary by account and region during the transition. The most important thing to watch is how Google explains the two-month retention period once the update reaches more users.
There are four open questions worth tracking:
- How much user data reviewers can access under the new retention policy.
- Whether the longer storage period changes legal or regulatory exposure.
- How many users receive the default-setting change, and how quickly rollout finishes.
- Whether Google gives a fuller explanation for shifting from 72 hours to about two months.
It will also be worth watching whether Google further clarifies the difference between chats stored for safety review, chats used for feedback, and chats reviewed by service providers. Those distinctions may sound technical, but they determine how much of a user’s Gemini conversation stays in Google’s systems and for how long.
The bigger takeaway
Google’s Gemini update is a privacy improvement in one sense and a retention expansion in another. The company is finally separating chat history from AI training consent, which puts Gemini closer to the competitive norm. But it is also preserving more data for longer, even when users say they do not want their chats used to train the model.
That combination captures the current state of consumer AI privacy: companies want to offer clearer controls, but they still want enough data to maintain safety, performance, and service quality. Gemini’s new policy does both at once — and that is exactly why it matters.
“temporary conversations are not used to train AI models”
“around two months”
Those two lines, taken together, summarize the tradeoff Google is making. The model-training side is cleaner than before. The retention side is not.
| Milestone | Date | What happened |
|---|---|---|
| Older Gemini policy | Oct. 26, 2026 | Privacy hub reportedly still showed 72-hour retention |
| Archived hub reference | Nov. 12, 2026 | TechCrunch said archived page still showed 72 hours |
| Privacy hub update | Nov. 16, 2026 | Notebookcheck said Gemini Apps Privacy Hub was updated |
| Public reporting | Nov. 17, 2026 | TechCrunch and others reported the privacy changes |
| Rollout window | Coming weeks | Google is rolling out the new settings globally |
Frequently asked questions
What changed in Gemini’s privacy settings?
Google is separating Gemini chat history from AI training consent, so users can keep history on without allowing training use. The company is also extending retention for temporary and history-off chats to about two months, up from 72 hours.
Does turning off Gemini history stop Google from storing chats?
No. Turning history off no longer means immediate deletion. According to the reporting, Google will still keep those chats for around two months for safety, security, and service-related purposes, although temporary conversations are not used to train AI models.
How does Gemini compare with ChatGPT and Claude on privacy?
Gemini is moving closer to ChatGPT and Claude by separating conversation history from model-training permission. Crypto Briefing said those competitors already offer that split, which makes Google’s old Keep Activity setup look more restrictive and less flexible.
Will feedback still matter if training is turned off?
Yes. Crypto Briefing reported that if a user rates a response or sends chat feedback, that interaction can still support model improvement. Notebookcheck also said feedback while History is off can collect the last 24 hours of chats, including attachments.









