Skip to content
ContentLora

    Tip: press / anywhere to search.

    Explainer

    How AI chatbots use your data: training, opt-outs and retention

    OpenAI, Anthropic and Google each say in their privacy pages that they can use consumer chats to improve their AI models, and each describes a setting that controls this.[1][2][3] Under the companies' stated policies, retention ranges from 72 hours for some Gemini chats to five years for Claude users who allow training.[4][5] Microsoft says Microsoft 365 Copilot work data is not used to train foundation models.[6] Italy's data protection authority fined OpenAI 15 million euros in 2024, but a Court of Rome judgment published on 18 March 2026 upheld the appeal against that decision; in 2025 the US FTC ordered seven chatbot companies to report on how they use conversation data.[7][8]

    Editor reviewedUpdated Explainer19 claims · 100% primaryArtificial intelligence, Tech policy
    Reading level: 1 passage differs; the rest is the same for everyone.

    This guide summarises each company’s own published policy as of the dates given (all pages retrieved on 7 November 2026). Policies change often and differ by country and plan, and each company’s current privacy page is the authoritative version. It is not legal advice. For the laws behind these rules, read our explainer on data privacy laws.

    The companies’ policies describe storing what users write to their chatbots and, in some cases, reusing it. The main questions are whether your chats help train future models, whether you can switch that off, and how long the chats are kept. The answers differ by company and by plan.[1][9][4]

    Training on your chats

    Training means a company uses past conversations to improve its next model. OpenAI’s February 2026 privacy policy says it may use your chats this way and that you can choose whether it does.[1] Anthropic’s August 2025 update says Claude users choose whether to share chats for training.[9] Google’s Gemini privacy hub says it uses some Gemini data to improve its AI when a setting called Keep Activity is on.[3] Microsoft’s privacy statement, last updated September 2026, says it may use data to train its AI models.[10]

    OpenAI’s privacy policy, updated February 6, 2026, says it may use user content to improve its services, for example to train the models behind ChatGPT, and that users can choose whether their content is used to improve and train its models.[1] Anthropic’s August 28, 2025 update applied to Claude Free, Pro and Max, including Claude Code on those accounts. Existing users had until October 8, 2025 to choose, and services under its Commercial Terms and the API were excluded.[9] Its privacy policy effective September 10, 2026 says inputs and outputs may be used for training unless the user opts out in account settings.[2]

    Google’s Gemini Apps Privacy Hub (retrieved 7 November 2026) says Gemini Live transcripts and uploads are used to improve its services, including AI models, when Keep Activity is on, and that Gemini Apps Activity lets users control whether data improves Google AI.[3] Microsoft’s privacy statement, last updated September 2026, says Copilot conversation data is used to provide and improve Copilot, and that Microsoft may use data to develop, train and fine-tune AI models, including large language models.[10]

    How long chats are kept

    Company (policy)What the policy saysClaim
    OpenAI (privacy policy, February 6, 2026)Deleted data removed within 30 days, with exceptions; Temporary Chats deleted within 30 days[11][12]
    Anthropic (consumer-terms update, August 28, 2025)Five years if you allow training; otherwise 30 days[5]
    Google Gemini (Privacy Hub, retrieved November 7, 2026)18 months by default; 72 hours with Keep Activity off; human-reviewed chats up to three years[4][13]

    OpenAI says deleted personal data is removed within 30 days unless it must be kept longer, or it has already been de-identified for model improvement.[11] Google says chats reviewed by humans are kept for up to three years, disconnected from the account, and are not deleted when you delete your activity.[13] Even with Keep Activity off or in a temporary chat, Google still uses chats to respond and to protect its users and the public, with help from human reviewers.[13]

    Opt-outs and private modes

    These are the settings each company describes in its own policy; this guide does not test them.

    • ChatGPT: OpenAI says users can choose whether their content is used to train its models. It says Temporary Chats do not appear in history, are not used to improve its models and are deleted within 30 days unless kept for safety or legal reasons.[1][12]
    • Claude: Anthropic’s 2026 privacy policy says users can opt out of training in their account settings, and its 2025 update said users who do not opt in keep the 30-day retention period.[2][5]
    • Gemini: Google says temporary chats and chats with Keep Activity off are kept with the account for 72 hours, and that users can change auto-delete from the 18-month default to 3 or 36 months or no auto-deletion.[4]
    • Meta AI: Meta said in October 2025 that tools such as Ads Preferences let users adjust the content and ads they see, and that from December 16, 2025 it would use AI interactions to personalise content and ads in most regions.[14]

    Work accounts and ads

    Business products are usually covered by separate contracts. Anthropic’s 2025 change did not apply to Claude for Work, Claude for Government, Claude for Education or the API.[9] Microsoft’s Microsoft 365 Copilot privacy documentation says prompts, responses and Graph data are not used to train foundation models.[6] Copilot for consumers is different: Microsoft’s privacy statement (last updated September 2026) says ads may relate to your Copilot conversations.[15]

    Meta said it would use interactions with Meta AI to personalise content and ads from December 16, 2025. It said it would not use conversations about religion, sexual orientation, politics, health and other sensitive topics to show ads.[14]

    What regulators have done

    Italy’s data protection authority, the Garante, acted against ChatGPT in 2023. On March 31 that year it ordered an immediate temporary limit on OpenAI’s processing of Italian users’ data.[16] On December 20, 2024 it announced a 15 million euro fine and a six-month information campaign. It found that OpenAI had processed personal data to train ChatGPT without first identifying an appropriate legal basis, had not reported a March 2023 data breach and lacked age checks. The Garante later said the Court of Rome, in judgment No. 4153/2026 published on March 18, 2026, upheld the appeal lodged against that decision.[7][17]

    In the US, the Federal Trade Commission ordered seven companies, including Alphabet, Meta, OpenAI, Snap and xAI, to report on their companion-style chatbots on September 11, 2025. The orders asked how they use or share personal information from users’ conversations.[8] The FTC said the study focuses on risks to children and teens and has no specific law-enforcement purpose.[18]

    In Europe, the AI Act’s transparency rules require that people be told when they are talking to a chatbot.[19] For the wider picture, read AI regulation around the world and data privacy laws explained.

    OpenAI

    OpenAI is the US AI company behind ChatGPT and the GPT models. Its structure, funding, products and what changed in 2026.

    More in AI

    Questions readers ask

    Does ChatGPT train on my conversations?

    It can. OpenAI's privacy policy says it may use content you provide to train the models that power ChatGPT, and that you can choose whether your content is used this way. Temporary Chats are not used to improve its models.[1][12]

    Does Claude train on my chats?

    Since Anthropic's 2025 consumer-terms update, users of Claude Free, Pro and Max choose whether their chats are used for training. Its 2026 privacy policy says it may use inputs and outputs for training unless you opt out in your account settings. Business and API use is covered by separate commercial terms.[9][2]

    How long does Gemini keep my chats?

    Google says Gemini activity is auto-deleted after 18 months by default, which you can change to 3 or 36 months or no deletion. Chats with Keep Activity off are kept for 72 hours, and chats reviewed by humans can be kept for up to three years.[4][13]

    Does Meta use my Meta AI chats for ads?

    Meta said that from December 16, 2025 it would use interactions with its AI to personalise content and ads in most regions, but not conversations about sensitive topics such as religion, politics or health.[14]

    Is my work data used to train Microsoft Copilot?

    Microsoft says prompts, responses and data accessed through Microsoft Graph in Microsoft 365 Copilot are not used to train foundation models.[6]

    Has any regulator fined a chatbot company over data?

    Italy's Garante announced a 15 million euro fine against OpenAI in December 2024, partly for training ChatGPT on personal data without an appropriate legal basis. The Garante says the Court of Rome, in judgment No. 4153/2026 published on March 18, 2026, upheld the appeal lodged against that decision.[7][17]

    Sources

    · 19 claims from 9 publishers

    Each numbered claim is a statement we checked against the sources listed with it. Tap a number in the text to see its sources in place.

    1. [1]

      OpenAI's privacy policy, updated 6 February 2026, says it may use content users provide to improve its services, for example to train the models that power ChatGPT, and that users can choose whether their content is used to improve and train its models. confirmedas of 2026-11-07

    2. [2]

      Anthropic's privacy policy effective 10 September 2026 says it may use users' inputs and outputs to train and improve its AI models unless they opt out in their account settings. confirmedas of 2026-11-07

    3. [3]

      Google says that when Keep Activity is on, transcripts of Gemini Live chats and uploads shared with Gemini Live are used to improve Google services, including AI models, and that Gemini Apps Activity lets users control whether their data is used to improve Google AI. confirmedas of 2026-11-07

    4. [4]

      Google's Gemini Apps Privacy Hub says Gemini activity is auto-deleted after 18 months by default, which users can change to 3 or 36 months or no auto-deletion, and that temporary chats and chats made with Keep Activity off are kept with the account for 72 hours. confirmedas of 2026-11-07

    5. [5]

      Anthropic said it would keep consumer chat data for five years for users who allow it to be used for model training, while users who do not opt in keep the existing 30-day retention period. confirmedas of 2025-08-28

    6. [6]

      Microsoft says prompts, responses and data accessed through Microsoft Graph in Microsoft 365 Copilot are not used to train foundation large language models, including those used by Microsoft Copilot. confirmedas of 2026-11-07

    7. [7]

      On 20 December 2024 the Garante announced a 15 million euro fine against OpenAI and ordered a six-month information campaign, finding that OpenAI had not notified it of a March 2023 data breach, had processed users' personal data to train ChatGPT without first identifying an appropriate legal basis, had breached transparency obligations and lacked age verification. The Garante later noted that judgment No. 4153/2026 of the Court of Rome, published on 18 March 2026, upheld the appeal lodged against that decision. confirmedas of 2026-11-07

    8. [8]

      On 11 September 2025 the US Federal Trade Commission issued 6(b) orders to seven companies running consumer AI chatbots - Alphabet, Character Technologies, Instagram, Meta, OpenAI, Snap and xAI - seeking information including how they use or share personal information obtained through users' conversations. confirmedas of 2025-09-11

    9. [9]

      On 28 August 2025 Anthropic updated its consumer terms so that users of Claude Free, Pro and Max, including Claude Code on those accounts, choose whether their chats are used for model training, with existing users given until 8 October 2025 to decide; the change did not apply to services under its Commercial Terms or API use. confirmedas of 2025-08-28

    10. [10]

      Microsoft's privacy statement, last updated in September 2026, says it uses Copilot conversation data to provide and improve Microsoft Copilot and may use personal data to develop, train and fine-tune its AI models, including large language models. confirmedas of 2026-11-07

    11. [11]

      OpenAI says that once a user deletes personal data it removes it from its systems within 30 days, unless it must keep it longer or the data has already been de-identified for model improvement. confirmedas of 2026-11-07

    12. [12]

      OpenAI says ChatGPT Temporary Chats do not appear in history, are not used to improve its models, and are automatically deleted within 30 days unless they must be kept for safety or legal reasons. confirmedas of 2026-11-07

    13. [13]

      Google says that even with Keep Activity off or in temporary chats it uses chats to respond and to protect its users and the public, including with help from human reviewers, and that chats reviewed by humans are kept for up to three years, disconnected from the account, and are not deleted when users delete their activity. confirmedas of 2026-11-07

    14. [14]

      Meta said on 1 October 2025 that from 16 December 2025 it would use people's interactions with its AI to personalise the content and ads they see, in most regions, while not using conversations about sensitive topics such as religion, sexual orientation, politics or health to show ads, and that users can adjust the content and ads they see with tools such as Ads Preferences. confirmedas of 2025-10-01

    15. [15]

      Microsoft's privacy statement says ads may be shown that relate to a user's current interaction, such as their Copilot conversations, including files they share. confirmedas of 2026-11-07

    16. [16]

      On 31 March 2023 Italy's data protection authority, the Garante, ordered with immediate effect a temporary limitation on OpenAI's processing of Italian users' data, citing unlawful collection of personal data and the absence of age verification. confirmedas of 2023-03-31

    17. [17]

      The Garante says its decision against OpenAI (No. 755 of 2 November 2024) was temporarily removed from its website after judgment No. 4153/2026 of the Court of Rome, published on 18 March 2026, upheld the appeal against it. confirmedas of 2026-11-07

    18. [18]

      The FTC said the 6(b) orders focus on how the companies measure, test and monitor potential negative impacts of their chatbots on children and teens, and that 6(b) studies do not have a specific law enforcement purpose. confirmedas of 2025-09-11

    19. [19]

      Under the AI Act's transparency rules, people must be made aware when they are interacting with an AI system such as a chatbot, and certain AI-generated content, namely deep fakes and text published to inform the public on matters of public interest, must be clearly labelled. confirmedas of 2026-11-07

    Revision history (1)
    1. Page created.

    Created Nov 7, 2026. Last reviewed by an editor on Nov 8, 2026. Next scheduled review: Feb 7, 2027.

    Cite this page

    "How AI chatbots use your data: training, opt-outs and retention." ContentLora, updated Nov 7, 2026. https://contentlora.com/explain/how-ai-chatbots-use-your-data

    Spotted an error? Suggest a correction or emailcorrections@contentlora.com.