Privacy by Editorial Team

How to Delete Your Data From ChatGPT, Claude, and Gemini

The exact steps to delete chats and accounts from ChatGPT, Claude, and Gemini, and what each one keeps on its servers after you hit delete.

Hitting the trash icon next to a chat in ChatGPT, Claude, or Gemini does not delete the conversation from the company’s servers. It removes the conversation from your view. Each of the three main consumer AI assistants treats “delete” as a multi-step process with a waiting period, several pieces of personal information that persist past that period, and a separate path for closing the account itself. This guide walks through each one, what survives deletion, and what your legal rights actually require a company to do. (For what each assistant collects in the first place, see our source-by-source privacy audit of eight popular AI assistants.)

TL;DR

  • “Delete” means delete from your view first. Anthropic and OpenAI describe a back-end deletion window of about 30 days; Google keeps chats for 72 hours even with activity saving off. Copies can then persist for safety, legal, or feedback reasons.
  • Claude and Gemini both document copies that survive deletion: Anthropic keeps rated conversations as Feedback and uses safety-flagged conversations, and Google keeps human-reviewed chats for up to three years, disconnected from the account.
  • Anthropic (Claude) lets you delete individual chats, and its Consumer Terms say you can close the account by contacting [email protected]. Anthropic deletes individual chats from its back-end within 30 days.
  • Google (Gemini) gives you the sharpest retention number (72 hours when Keep Activity is off; auto-delete after 3, 18, or 36 months when Keep Activity is on). Human-reviewed copies are kept for up to 3 years and disconnected from your Google account.
  • OpenAI (ChatGPT) runs on a 30-day deletion window for backups, with carve-outs for legal obligations such as litigation holds. Past chats already in a training set are not retroactively removed.
  • Under EU GDPR Article 17, you have a right to erasure. Providers may refuse where a legal obligation, public interest, or the establishment of legal claims requires keeping the data.

What “delete” actually means in three places

The conversation you see in your chat sidebar is a frontend view of a record that lives in a backend. The three companies each treat that record differently:

  • Anthropic. Anthropic’s Privacy Policy says individual conversations “will be removed immediately from your conversation history and automatically deleted from our back-end within 30 days.” The 30 days is a worst-case. Training-flagged or feedback-flagged copies are kept on a different track.
  • Google. Gemini Apps Privacy Hub explains that activity is auto-deleted after 18 months by default (configurable to 3 or 36 months, or off). When Keep Activity is off, Gemini retains the conversation “with your account for 72 hours” for context. Human-reviewed chats are kept for up to three years and disconnected from your account.
  • OpenAI. OpenAI’s privacy documentation describes a 30-day retention window for backups after deletion. Conversations remain in your account until you delete them; deletion removes them from view, with a 30-day backup window for safety and legal reasons.

The short-term windows differ (30 days at Anthropic and OpenAI, 72 hours at Google with activity off). What matters more is what each company keeps on top of that window.

Claude: delete a chat, delete an account

For a single conversation, click the conversation in the sidebar, open the menu, and select Delete. The Anthropic Privacy Policy is explicit: “You also are able to delete individual conversations… which will be removed immediately from your conversation history and automatically deleted from our back-end within 30 days.”

To close the account itself, the Anthropic Consumer Terms (effective October 8 2025) state that “You may close your Account at any time by contacting us at [email protected].” On termination, the Terms say “we may at our option delete any Materials or other data associated with your Account.” That is a permission to delete, not a commitment to a deletion date.

Two Anthropic carve-outs to know:

  1. Feedback retention. “If you rate an Output in response to an Input - for example, by using the thumbs up/thumbs down icon - we may store the entire related conversation as part of your Feedback.” That conversation persists as Feedback regardless of whether you delete the original chat.
  2. Safety-flag retention. The Anthropic Privacy Policy states that even when a user opts out of training, “we will use Inputs and Outputs for model improvement when… your conversations are flagged for safety review to improve our ability to detect harmful content, enforce our policies, or advance AI safety research.” Flagged content is de-identified for trust and safety models, but Anthropic notes it may re-identify the content “to enforce our Terms of Service or Usage Policy with the responsible user if necessary.”

For GDPR data-subject requests (access, deletion, rectification), Anthropic’s Privacy Policy states it “will respond within one calendar month… we may extend that period by up to a further two months.”

Gemini: settings, activity page, and the three-year archive

Gemini gives the user more direct control than the other two. The path is browser or mobile, then settings.

On the web:

  1. Open gemini.google.com.
  2. Click Settings & help, then Activity.
  3. Or go straight to myactivity.google.com/product/gemini.
  4. Use the Delete menu above the activity list to remove the last hour, last day, a custom date range, or all time; single days and single items can be deleted too.
  5. Change auto-delete under “Auto-delete” to 3 months, 18 months, 36 months, or Don’t auto-delete.

The Gemini Apps Activity help page is explicit: “By default, your Gemini Apps activity older than 18 months is auto-deleted. You can turn off Auto-delete or change the auto-delete period to 3 or 36 months.”

The same help page has separate instructions for the Android and iPhone apps.

The retention rule depends on which mode the user is in:

  • Keep Activity on (the default). Activity is retained with the account and auto-deleted at the chosen interval (3, 18, 36 months, or never).
  • Keep Activity off. Google says these chats are “retained with your account for 72 hours” so Gemini can respond, process feedback, and protect Google, its users, and the public; they are not used to train its models unless you send feedback.
  • Temporary Chat. Temporary chats are not used to train Google’s models, but they are also retained with your account for 72 hours.

Two things persist past activity deletion, per the Gemini Apps Privacy Hub:

  1. Human-reviewed chats are “retained for up to three years” and “disconnected from your account before being sent to service providers.” Deleting your Gemini activity does not delete the human-review copy.
  2. Reviewed feedback “and related data are retained for up to 3 years, disconnected from your Google Account.”

Chats imported from other AI platforms are saved in Gemini Apps Activity like any other chat, and can be deleted individually or in bulk from there.

To delete the Google account entirely (not just Gemini), that is done through Google Account → Delete your Google Account, not through the Gemini settings.

ChatGPT: data controls, the privacy portal, and what “improve model for everyone” hides

The 30-day window that covers ChatGPT is the same industry standard. The user-facing path lives under Settings → Data Controls.

For a single chat: open the chat, hover or tap the three-dot menu, choose Delete. OpenAI’s documentation describes the result as removing the conversation from your view while a 30-day backup window holds the record for safety and legal reasons.

For training, OpenAI documents the opt-out on its help page under “How do I opt out of my data being used to train future models?” The toggle lives at Settings → Data Controls → “Improve model for everyone.” When enabled (default for new accounts), OpenAI may use your conversations to train future models. When disabled, future conversations are not used. Two footguns apply: toggling off does not retroactively remove past chats already used for training, and OpenAI may still review flagged conversations regardless of the toggle.

To delete the account: Settings → Data Controls → Delete account → confirm with the “Permanently delete my account” button. OpenAI’s documented retention policy is that data is generally purged within 30 days unless a legal or safety obligation requires longer retention. The OpenAI Help Center describes a 30-day backup window for ordinary deletion, and a separate path through privacy.openai.com for GDPR data-subject requests.

Three ChatGPT-specific retention carve-outs that matter:

  1. Training carve-out. Past conversations already used for training are not retroactively removed when you toggle training off. OpenAI’s help page says the control affects future data only.
  2. Backup window. OpenAI’s privacy documentation states that backups may retain deleted data for up to 30 additional days unless OpenAI is required to retain it longer for security or legal obligations.
  3. Litigation. Court orders can reach chat logs. In the NYT v. OpenAI copyright case, Magistrate Judge Ona T. Wang ordered OpenAI in November 2025 to produce a de-identified sample of 20 million ChatGPT conversation logs, and U.S. District Judge Sidney Stein affirmed that order on January 5, 2026.

For EU/UK users, the OpenAI EU Privacy Policy describes the company as a data controller under GDPR with the standard data-subject rights. Requests can be submitted through privacy.openai.com.

Under EU GDPR Article 17, the right to erasure (the “right to be forgotten”) gives a data subject the right to obtain erasure of personal data “without undue delay where one of the following grounds applies”: the data is no longer necessary for the purpose it was collected, consent is withdrawn, the data subject objects, the data was unlawfully processed, there is a legal obligation to erase, or the data was collected from a child in connection with an online service.

Article 17(3) lists the exceptions where the right does not apply: freedom of expression and information; compliance with a legal obligation or a task in the public interest; public health; archiving in the public interest, scientific or historical research, or statistics; and the establishment, exercise, or defence of legal claims. The legal-obligation and legal-claims exceptions are the ones that line up with the retention carve-outs described above.

Outside the EU, the picture is uneven. US federal law has no general right to erasure. California (CCPA/CPRA) gives residents the right to delete personal information collected about them, with business-disclosed exceptions for security, fraud prevention, and legal compliance. Other US states have similar but distinct rules.

The Bottom Line

Hitting Delete in ChatGPT, Claude, or Gemini removes the conversation from your view; the rest depends on which company, which mode, and which flag the conversation carried. Anthropic gives the cleanest single-chat deletion (immediate from history, 30 days from back-end) but reserves the right to keep feedback-flagged and safety-flagged content. Google gives the sharpest retention number (72 hours when Keep Activity is off, auto-delete at 3/18/36 months when on) but holds human-reviewed and feedback data for up to three years. OpenAI describes a 30-day backup window but reserves the right to keep data longer for legal obligations, and says its training toggle affects future chats only. If you live in the EU or UK, GDPR Article 17 gives you a legal right to demand erasure, subject to the exceptions in Article 17(3). Read the retention number. Read the carve-outs. Then decide whether “delete” matches the deletion you actually wanted.