Gemini Now Keeps Temporary Chats for 2 Months. Check Your Team.
Google is extending how long it keeps temporary Gemini chats from 72 hours to about two months. If your team pastes customer data, financials or code into consumer AI tools, the retention window just changed. Here is what each assistant keeps, and a one-page AI policy a five-person startup can adopt this week.

Somebody on your team pasted something into an AI chatbot this week. Maybe a customer list. Maybe last quarter's numbers. Maybe the contract a prospect sent over for redlines.
If they used a temporary chat in Google's Gemini app, they probably assumed it was gone in three days. That assumption is about to expire.
TechCrunch reported on November 17 that Google is extending how long it keeps temporary Gemini chats, and chats made with history turned off, to about two months. The previous window was 72 hours, according to an archived version of Google's Gemini Privacy Hub dated November 12. The change rolls out globally in the Gemini app "over the coming weeks."
None of this means Google is doing something sinister with your data. It does mean the rules your team is quietly relying on have changed, and most small startups have no written rules at all. Let's fix that.
What exactly did Google change?
Two things: how long throwaway chats are kept, and how the privacy settings are split.
On retention, Google says the longer window is meant to "improve safety and security as AI systems become more capable." Its post says chats are kept to respond to users, process feedback, and protect its services and users. The Privacy Hub also lists potential system failures as a reason. TechCrunch notes Google didn't say what prompted the change or whether it affects how conversations are accessed or reviewed.
On settings, Google is splitting the single "Keep Activity" toggle into two:
- One controls whether conversations are saved to your chat history
- One controls whether conversations can be used to train Google's generative AI models
If you'd turned Keep Activity off, both new settings start off. If you'd left it on, chat history stays on, and you get to decide separately about training. The Privacy Hub says temporary conversations aren't used for training.
For context, temporary chats only arrived in August 2025. Google's launch post at the time said they'd be "kept for up to 72 hours." So a feature people chose specifically for its short memory is getting a much longer one.
Why should a five-person startup care about retention?
Because retention is exposure. The longer data sits somewhere, the more chances something goes wrong with it: a breach, a legal request, a reviewer reading a sample, a setting someone flips by mistake.
And it's not hypothetical. In May 2023, Bloomberg reported that Samsung banned employees from using ChatGPT and other AI chatbots after an engineer accidentally uploaded sensitive internal source code. Samsung's worry, as reported, was that data shared with these services sits on outside servers, can be hard to retrieve or delete, and could end up shown to other users.
Samsung has a security team. You probably have a shared Google Drive and good intentions.
Here's the other piece most founders miss. Back in February 2024, TechCrunch reported that Google's support page said human reviewers read and label some Gemini conversations, and those reviewed conversations can be kept for up to three years. The same page carried a plain warning: "Please don't enter confidential information in your conversations."
That warning is the whole policy, really. The rest is helping your team follow it.
How long do the big assistants keep temporary chats?
Roughly a month to two months for the main consumer apps, with exceptions for legal and safety reasons.
Based on TechCrunch's November 17 comparison:
| Assistant | Temporary or history-off chats |
|---|---|
| Google Gemini | About two months (was 72 hours) |
| OpenAI ChatGPT | Up to 30 days for temporary chats |
| Anthropic Claude | 30 days by default for incognito chats |
All three companies may keep some data longer for legal or safety reasons, or to investigate policy violations.
Two cautions. First, these are consumer app policies. TechCrunch didn't say whether Gemini's change treats business accounts differently, so check the terms for whatever plan your company actually pays for instead of assuming. Second, policies change. This one changed in about 15 months. Put a calendar reminder to recheck twice a year.
Your AI co-founder is ready when you are.
Foundra turns everything in this article into an actual plan. Validation, customers, pricing, launch. In one place, in your voice, in an afternoon.
Get started→$39/month. Cancel anytime.
What should never go into a consumer AI chat?
Anything you'd be uncomfortable seeing in a breach notice, a lawsuit, or a competitor's inbox.
For an early startup, that list usually looks like this:
- Customer personal data: names with emails, phone numbers, addresses, payment details
- Unreleased financials: bank balances, burn, runway, and investor updates before they're sent
- Your cap table and any term sheet you're negotiating
- Source code containing secrets, like API keys, tokens or database credentials
- Employee matters: performance issues, salaries, health or leave details
- Anything under NDA, including a partner's or prospect's documents
Notice what's not on the list. Brainstorming taglines, cleaning up a public blog post, explaining a concept, writing a generic email template. Most of the value people get from these tools never needed sensitive data in the first place.
The trick is the middle ground. Want help with a pricing model? Swap real customer names for "Customer A." Want feedback on an investor update? Remove the cash figure and keep the structure. You lose almost nothing, and the risk drops to near zero.
How do you write an AI policy in one afternoon?
Keep it to one page, write it in plain language, and make the safe path the easy path.
Step 1: Pick approved tools. Name the one or two AI assistants the company pays for, on business plans where you've read the data terms. Personal accounts for company work are out.
Step 2: Copy the never-paste list from the section above into the doc. Add anything specific to your industry.
Step 3: Set the defaults. For each approved tool, write down the exact settings (training off, history on or off) and ask everyone to screenshot theirs once.
Step 4: Keep the source of truth somewhere you control. Your plan, your model and your customer records should live in tools your team owns, whether that's Google Docs, Notion, or a planning tool like Foundra, with only the needed slice pasted into an AI chat when a question calls for it.
Step 5: Say what to do after a mistake. Someone will paste the wrong thing eventually. Tell them to delete the chat, tell you, and not panic. A team that hides mistakes is more dangerous than one that makes them.
Step 6: Review it twice a year. Today's news is the reason why.
I've seen teams skip this because it feels like big-company bureaucracy. It's the opposite. A one-page rule written at five people is far easier than unwinding habits at fifty.
What about customers who ask about your AI use?
Expect the question, especially if you sell to businesses. Security questionnaires increasingly ask which AI tools touch customer data and under what terms.
Having the one-page policy means you can answer in a sentence: which tools, which plans, which data is excluded. That's a sales asset. A founder who can say "customer data never goes into consumer chatbots, here's our policy" sounds a lot more trustworthy than one who has to go ask the team.
Key takeaways
- Google is extending retention for temporary Gemini chats and history-off chats from 72 hours to about two months, rolling out globally over the coming weeks.
- Keep Activity is being split into two settings: one for chat history and one for AI training.
- Per TechCrunch, ChatGPT temporary chats can be kept up to 30 days, and Claude incognito chats 30 days by default.
- Google has long warned users not to enter confidential information, and reviewed Gemini chats have been kept for up to three years.
- Samsung banned AI chatbots in 2023 after an engineer uploaded internal source code.
- A one-page AI policy (approved tools, a never-paste list, default settings, a mistake plan) protects a small team and helps close deals.
Frequently asked questions
How long does Gemini keep temporary chats now? About two months, up from 72 hours, according to TechCrunch's November 17 report. The change is rolling out globally in the Gemini app over the coming weeks.
Are Gemini temporary chats used to train AI models? Google's Privacy Hub says temporary conversations are not used to train its models.
Does this change apply to Google Workspace business accounts? TechCrunch didn't say. Check the data terms for your specific plan rather than assuming consumer rules apply.
Is it safe to paste code into an AI assistant? Generic code is usually low risk. Code that contains API keys, credentials or proprietary logic shouldn't go into a consumer chat. Strip secrets first, or use an approved business plan.
Do startups really need an AI usage policy? Yes, and it can be one page. It sets clear limits, makes security questionnaires easier, and tells people what to do after a mistake.
Sources
- Google to keep Gemini chats for two months, up from 72 hours, even with history off, TechCrunch, November 17, 2026
- Gemini adds Temporary Chats and new personalization features, Google Blog, August 13, 2025
- Google saves your conversations with Gemini for years by default, TechCrunch, February 8, 2024
- Samsung bans ChatGPT among employees after sensitive code leak, Forbes Australia (via Bloomberg), May 2, 2023
You just read the theory. Ready to build the thing?
Foundra is your AI co-founder. It turns an idea into a validated business plan, a go-to-market, and your first 10 customers. In an afternoon, not a semester.
$39/month. Cancel anytime. Works in 20 languages.