How long does ChatGPT keep your chats?

ChatGPT keeps your chats until you delete them. There is no automatic expiry on a saved conversation: OpenAI's privacy policy, updated February 6, 2026, lists ChatGPT conversations as information it retains until you delete it. Once you delete a chat, or the whole account, OpenAI says it removes the data from its systems within 30 days, unless it has to keep it longer for safety, legal or financial reasons. Temporary Chats are deleted automatically within 30 days. The other big assistants work on different clocks: Gemini deletes activity after 18 months by default, Copilot keeps the last 18 months of conversation history, and Claude keeps chats for five years if you let Anthropic train on them.

How long each AI chatbot keeps your chats

These are the consumer accounts, free or paid, as each company currently words its policy. Work, school and developer accounts run on separate contracts and are covered further down.

Two patterns stand out. Thirty days is the industry's standard purge window after you press delete, and it appears in almost every row. The length of time a chat survives if you do nothing, though, ranges from "until you act" at OpenAI and Grok to 18 months at Google and Microsoft, the only two on the list that cap your history without being asked.

Does ChatGPT delete your chats after 30 days?

No, and this is the misunderstanding behind most searches on the subject. The 30 days in OpenAI's policy is a purge window, not a lifespan. It starts when you delete something, or when you open a Temporary Chat, and it describes how long the copy lingers on OpenAI's side after it has vanished from yours. A conversation you leave in the sidebar has no end date at all.

The confusion persists because 30 days turns up in three places at once: the deletion window, Temporary Chat, and the safety copy OpenAI may keep of a temporary conversation. OpenAI's help page on Temporary Chat says it "may keep a copy of a temporary chat for up to 30 days for safety purposes", and that if you save a temporary chat as a regular one, it then follows the retention rules for saved chats. Read quickly, all of that sounds like a monthly cleanup. It is not one.

There is a second qualifier in the policy that matters if model training is switched on. OpenAI's 30 day promise does not cover content that "has already been de-identified and disassociated from your account" because you allowed it to be used to improve the models. Deleting a chat removes your copy, and it does not reach back into a training set. What deleting AI chat data actually removes explains why no provider can pull a conversation back out of a model it has already trained.

That separate store is easy to miss, and how chat history and memory differ shows where each one lives in the settings.

What the New York Times lawsuit changed

For several months in 2025, OpenAI's 30 day deletion promise was on hold. In May 2025, as part of the copyright case brought by The New York Times, OpenAI was ordered to preserve its ChatGPT logs, which meant chats users deleted were being kept rather than purged.

That obligation ended on September 26, 2025. As Engadget reported, federal judge Ona T. Wang filed an order on October 9, 2025 freeing OpenAI from having to "preserve and segregate all output log data that would otherwise be deleted on a going forward basis." Two things carried over. Logs already saved under the earlier order remain preserved, and OpenAI must keep holding data tied to ChatGPT accounts the Times has flagged.

So the answer that circulated in mid 2025, that OpenAI keeps everything forever, was true for a while and is not the rule now. A chat you delete today goes back on the 30 day clock. If you deleted something sensitive between May and late September 2025, though, the deletion did not work the way the policy promised at the time.

The copies that outlive deletion

Every policy above carries an exception for safety and legal reasons. A few providers put numbers on it, and those numbers are far longer than 30 days.

• Flagged by safety systems. Anthropic keeps the inputs and outputs of conversations flagged as a Usage Policy violation for up to 2 years, and the trust and safety classification scores for up to 7 years. Its Incognito chats are deleted within 30 days unless flagged in this way. OpenAI says it may keep content or accounts banned for policy violations for longer, without giving a period.
• Read by a human reviewer. Google states that Gemini chats reviewed by people are not deleted when you delete your activity. They are kept for up to three years, along with related data such as your language, device type, location information and feedback.
• Sent as feedback. When you rate a Claude response with a thumbs up or down, Anthropic keeps the data associated with that submission for 5 years.
• Under a legal hold. OpenAI's policy says data covered by a legal obligation, such as a lawful subpoena, is kept for the duration of that obligation. Grok's and Anthropic's policies name legal requirements as an exception too.

The practical effect is that the button you press decides the copy you can see, while what you typed decides how long the provider's copy lives. That is why the details you should never paste into a chatbot matter more than any retention setting.

Where these numbers stop applying

Everything in the table is for personal accounts. Business products sit on contracts with their own retention terms: Anthropic's five year change explicitly excludes Claude for Work, Claude for Government, Claude for Education and API use, and its commercial products delete inputs and outputs within 30 days by default. If ChatGPT or Copilot was given to you by an employer, the workspace administrator sets the rules, and whether your employer can read your AI chats covers who else has access. The difference between the consumer app, a business plan and the developer API is laid out in where your prompts go after you press send.

Copilot is also in transition. Microsoft released a new version of the Copilot app on August 18, 2026, and its privacy FAQ says the details there apply to the older version, with separate documentation for the new one. The 18 month figure is from Microsoft's own conversation history page, but treat the Copilot row as the one most likely to change.

Finally, this is a reading of published policies, not legal advice, and none of these companies is bound to keep the numbers it has today. In the EU and UK you can ask a provider directly what it holds about you and how long it will keep it, using the access and erasure rights explained here.

How to keep your chats for less time

1. Use the temporary mode for anything sensitive. ChatGPT's Temporary Chat, Claude's Incognito chats and Grok's Private Chat all put a 30 day ceiling on the copy, unless it is kept for safety or legal reasons. Gemini's temporary chats go after 72 hours.
2. Shorten Gemini's clock. Change auto-delete from the default 18 months to 3 months, or turn Keep Activity off, which still leaves 72 hours of retention.
3. Decide on training before you type. On Claude, allowing training is what moves retention from 30 days to five years. The steps for each service are in how to stop your chats being used to train AI models.
4. Delete chats you no longer need, then clear memory. On every service above, nothing short of deleting starts the 30 day purge.
5. Skip the feedback buttons on anything private. On Claude, a single thumbs up keeps the data from that rating for five years.

If the only acceptable retention period is none, the answer is not a setting but a model that never sends your text anywhere, which is what running AI locally on your own device gives you.