Guides › AI data retention periods
How Long Do AI Providers Keep Your Data? Retention Periods for ChatGPT, Claude, Copilot and Gemini
There is no single number. Thirty days is the figure that recurs: it is how long OpenAI keeps API abuse monitoring logs, how long a deleted ChatGPT or Claude conversation takes to leave back-end systems, and the retention Anthropic requires on its most capable models. Around that figure, saved conversations on business plans are kept for as long as your administrators decide, which on several platforms defaults to indefinitely, and knowledge files in projects stay until someone deletes them. This page lists the retention periods each provider documents for each plan, separates the four clocks that run on every platform, and covers what outlasts your settings.
- Saved content. Kept until deleted on consumer plans. Set by admins on business plans, with defaults that are often indefinite.
- Deleted content. Removed from back-end systems within 30 days on ChatGPT and Claude.
- Knowledge files. Project files, GPT knowledge, hosted vector stores and Notebook copies persist until deleted. They are the longest-lived data on each platform.
- Exceptions. Flagged content (up to 2 years at Anthropic), human-reviewed chats (up to 3 years at Google), feedback (5 years at Anthropic) and legal holds outlast any setting.
The Four Clocks
Every provider runs four separate retention clocks, and most confusion comes from quoting one as if it were another. The first is saved content: conversations, files and projects that stay because the product keeps history. The second is deletion lag: the time between a user deleting something and the provider removing it from back-end storage. The third is safety and abuse logging, which the provider keeps for its own monitoring regardless of what the user saves. The fourth is the exception clock: content that was flagged, reviewed, submitted as feedback or placed under a legal hold. A 30-day figure in a provider's documentation usually refers to the second or third clock and says nothing about the first.
Retention by Product and Plan
Each row paraphrases the provider's published position as of November 2026. The provider policy comparison covers training, residency and admin controls. This table is the retention column in full.
| Product and plan | Saved conversations | After deletion or with history off | Knowledge files and exceptions |
|---|---|---|---|
| ChatGPT Free, Plus, Pro | Kept until the user deletes them. | Deleted chats are scheduled for permanent deletion within 30 days. Temporary chats may be kept up to 30 days for safety. | Project and GPT knowledge files persist until deleted. |
| ChatGPT Enterprise, Edu, Healthcare | Set by workspace admins. | Removed from OpenAI systems within 30 days unless legally required. | Project and GPT knowledge files persist until deleted or removed by a workspace policy. Synced sources keep a separate index. |
| OpenAI API | Nothing is saved unless a stateful feature is used. Abuse monitoring logs are kept up to 30 days. | Assistants objects are removed 30 days after you delete them. | Files, vector stores and conversations stay until deleted. Stored responses are kept at least 30 days. Prompt cache data expires within 24 hours. |
| Claude Free, Pro, Max | Kept until the user deletes them. | Deleted from back-end storage within 30 days. | De-identified data may be kept up to 5 years if model improvement is allowed. Flagged content up to 2 years. |
| Claude Team, Enterprise | Kept in the product. Enterprise Owners can set a custom period of 30 days or more. The default is indefinite. | Deleted from back-end storage within 30 days. Incognito chats are deleted within 30 days. | Projects are retained indefinitely by default. Flagged content up to 2 years, safety classification scores up to 7 years. |
| Anthropic API | Conversation content is not retained by default, per the API documentation. Covered Models require 30 days. | Not applicable to stateless calls. | Files API until deleted or expired. Batch 29 days. Code execution containers up to 30 days. Flagged content up to 2 years. |
| Microsoft Copilot, personal account | 18 months by default. | Users can delete individual conversations or their whole history. | Uploaded files are stored for up to 18 months. |
| Microsoft 365 Copilot | Stored with other Microsoft 365 content under your Purview retention policies. | Users can delete their Copilot activity history. Retention policies decide the rest. | Source files stay in Microsoft 365 under their own retention and labels. |
| Gemini app, personal Google Account | Auto-deleted after 18 months by default. Can be set to 3 or 36 months, or to no auto-delete. | With Keep Activity off, and in temporary chats, conversations are held for 72 hours. | Chats seen by human reviewers are kept up to 3 years and are not removed when activity is deleted. |
| Gemini for Google Workspace | Set by admins: off, an auto-delete period such as 3 months, 18 months or 3 years, or indefinite. | With history off, new chats are still saved for up to 72 hours. | Gemini Notebook keeps its own copy of each Drive file added to it. Vault retention applies to supported Workspace data. |
Why Knowledge Base Content Lasts Longest
Conversation retention gets the attention, and knowledge content is where the long-lived copies are. A Claude project is retained indefinitely by default, and a chat inside a project follows the project's retention, not the chat setting. ChatGPT project files and custom GPT knowledge files persist until deleted or until a workspace policy removes them, and OpenAI notes that synced sources keep a separate index. On the OpenAI API, files and hosted vector stores are kept until deleted. On Anthropic's API, uploaded files are kept until deleted or expired. Gemini Notebook stores its own copy of each Drive file a user adds.
The practical effect is that a retention policy written for chats leaves the reference material untouched. A review should list each knowledge feature in use, who can add files to it and what removes them. The knowledge features guide covers the assistant side, and the page on zero data retention for RAG covers the API side.
What Outlasts Your Settings
Four kinds of content stay beyond the period you configure. Flagged content: Anthropic keeps inputs and outputs flagged by its trust and safety systems for up to two years, and classification scores for up to seven. Reviewed content: Google keeps consumer Gemini chats that human reviewers have seen for up to three years, disconnected from the account. Feedback: Anthropic stores a conversation submitted with a thumbs up or down for five years. Compliance records: Anthropic's Activity Feed retains data for six years.
Legal orders are the fifth. In 2025 a court order in copyright litigation required OpenAI to preserve consumer ChatGPT and API content that its policy would otherwise have deleted. OpenAI stated that the order did not apply to ChatGPT Enterprise, ChatGPT Edu or zero data retention API traffic, and has since said the obligation ended. A provider's deletion schedule is a default that other obligations can override. Content that never reached the provider in identifiable form is not exposed to any of them.
Anthropic's Two Statements on API Retention
Anthropic's sources describe API retention in two ways. Its commercial retention article says API inputs and outputs are deleted from its back end within 30 days of receipt or generation. Its API documentation says conversation content is not retained by default, with Covered Models held for 30 days. Read together, 30 days is the outer bound for standard API traffic, and your contract and the data retention setting in the Claude Console state what applies to your organization.
Setting Retention Deliberately
- List every place content is stored.Chats, projects, knowledge files, connectors, API state and your own logs.
- Take the period from your obligations.Records rules may require keeping some content and privacy rules may require deleting it. Decide per data category.
- Set chats and projects separately.On Claude Enterprise the project setting overrides the chat setting. On ChatGPT, knowledge files need their own policy.
- Cover the API.Keep sensitive data off stateful endpoints, or set an expiry on uploaded files.
- Match your own systems.Gateways and observability tools often keep prompts longer than the provider does.
- Review quarterly.Defaults change, and new features bring new stores.
If the requirement is no retention at all, the pages on ZDR on enterprise AI platforms and how to get ZDR cover what is available.
Where Hardshell Sits
Hardshell changes what the retained copy contains. It connects enterprise data to ChatGPT, Claude, Microsoft 365 Copilot, Gemini and model APIs without migrating it, enforces each user's existing permissions at query time, and substitutes stand-ins for real identifiers before content reaches the provider, so a conversation kept for 30 days or indefinitely holds stand-ins in place of customer, patient or account identifiers. Every retrieval is recorded under opaque IDs in your environment, under your own retention policy. Hardshell deploys as a cloud service, in a private cloud or as an offline bundle.
Frequently Asked Questions
How long does ChatGPT keep my data?
On Free, Plus and Pro, conversations stay until you delete them, and deleted chats are scheduled for permanent deletion within 30 days. On Enterprise, Edu and Healthcare, workspace admins set the period. On the API, abuse monitoring logs are kept up to 30 days, and files, vector stores and stored conversations stay until deleted.
How long does Claude keep conversations?
On consumer plans, until you delete them, then up to 30 days in back-end storage. On Claude Enterprise, indefinitely unless an Owner sets a custom period of at least 30 days, with projects retained indefinitely by default. Flagged content can be kept for up to two years.
How long does Microsoft Copilot keep chat history?
Personal Copilot stores conversation activity for 18 months by default. Microsoft 365 Copilot stores interactions with your other Microsoft 365 content, and your Purview retention policies decide how long they are kept.
How long does Gemini keep my chats?
The consumer Gemini app auto-deletes activity after 18 months by default, adjustable to 3 or 36 months or to no auto-delete. In Google Workspace, admins set the period or turn history off, which leaves a 72-hour hold.
Does deleting a chat delete it immediately?
It disappears from your history immediately. OpenAI and Anthropic both state that deleted conversations are removed from back-end systems within 30 days, and content that was flagged, human-reviewed, submitted as feedback or placed under legal hold can be kept longer.
What is the shortest retention an AI provider offers?
Zero data retention on the OpenAI and Anthropic APIs, for approved organizations and eligible endpoints. Among the assistants, Gemini for Workspace holds chats for up to 72 hours with history off, and Claude Enterprise documents a 30-day minimum.
Related guides
- What AI providers do with your data · Training, retention, residency and admin controls side by side.
- ZDR on ChatGPT Enterprise, Claude Enterprise, Copilot and Gemini · What the enterprise assistants offer in place of zero retention.
- Do AI providers train on your data? · The answer for each plan, with the exceptions.
Sources (accessed November 2026)
OpenAI, Enterprise privacy at OpenAI; Data controls in the OpenAI platform; Chat and file retention in ChatGPT; How we're responding to The New York Times' data demands. Anthropic, How long do you store my data?; How long do you store my organization's data?; Configure custom data retention controls for Enterprise plans; API and data retention. Microsoft Learn, Data, privacy, and security for Microsoft 365 Copilot; Microsoft Support, Privacy FAQ for Microsoft Copilot. Google, Generative AI in Google Workspace privacy hub; Gemini Apps Privacy Hub.