Who owns your AI conversations?

When you type into an artificial intelligence (AI) assistant, your words go to the provider's servers and are stored under its terms. Those terms, plus your settings, decide who can read the chat, whether it trains future models, how long it's kept and when it can be handed to someone else.

So instead of asking who owns the words, ask the practical questions: who can read them, who can reuse them, how long they last, and whether you can take them with you or get rid of them.

People at the company, in limited cases. Google says a subset of Gemini chats is reviewed by people, including trained service providers, and that reviewed chats are kept for up to 3 years, disconnected from your account, even if you delete your activity.³ Anthropic keeps the inputs and outputs of chats flagged for policy violations for up to 2 years.⁷

Your employer, on a work account. Microsoft says prompts and responses in Copilot Chat for work are logged, and IT administrators can search and audit them.⁸ On ChatGPT and Claude business plans, the organization, not the individual user, controls exports of your chats.¹⁹

Courts and authorities, through legal process. OpenAI's privacy policy says it may keep data longer when legally required, such as under a subpoena, and may share it with government authorities to comply with the law.¹⁰

This has happened. In 2025 a court order in a copyright lawsuit required OpenAI to preserve consumer ChatGPT conversations, including deleted ones; OpenAI says the order ended on September 26, 2025, and that it still holds a limited set of April to September 2025 data under legal hold.¹¹

ChatGPT (Free, Go, Plus and Pro): sharing for training is on by default, and you can turn off "Improve the model for everyone" in Data controls.¹ Rating a reply with a thumbs up or down can still send that whole conversation for training after you opt out; Temporary Chats aren't used for training.¹²

Claude (Free, Pro and Max): since 2025, Anthropic asks each user to choose whether chats can train its models.² If you allow it, de-identified chats can be kept for up to 5 years, and Incognito chats aren't used for training either way.⁷¹³

Gemini: the Keep Activity setting controls it. When it's on, chats are saved (auto-deleted after 18 months by default) and can be used to improve Google's AI; when it's off, Google keeps chats for 72 hours and doesn't use them for training unless you send feedback.³ Temporary chats are also kept for 72 hours and aren't used for training, and Google notes that its Gemini settings don't control the creation of anonymized data from your chats.³

Microsoft Copilot (personal account): Microsoft says that in the updated Copilot app available since August 18, 2026, your prompts, responses and files aren't used to train the core models behind its AI.⁴ If you haven't updated, the older terms apply, and chats can be used for training unless you opt out.¹⁴

Meta AI: Meta lists your interactions with its AI features, along with public posts and photos, among its training sources.¹⁵ It says it doesn't train on private messages with friends and family unless someone in the chat shares them with the AI.¹⁵

Business plans work differently. OpenAI doesn't train on ChatGPT Business, Enterprise, Edu or developer (API) data by default, and neither does Anthropic for Claude for Work or its API, though Anthropic may use chats that staff rate with a thumbs up or down unless an owner turns rating off.¹²¹⁶ Google says Workspace content isn't used to train models outside your organization without permission, and Microsoft says work-account Copilot Chat isn't used to train the core models behind its AI.¹⁷⁸

ChatGPT began testing ads on its Free and Go plans in the US on February 9, 2026.⁵ Ads can draw on your current chat and, if ad personalization is on, on past chats and memory; OpenAI says it doesn't share conversations with advertisers.⁵

Microsoft says people without a Microsoft 365 subscription may see ads in Copilot, and personalized ads can use chat history and saved memories unless you turn that off.¹⁸ Since December 16, 2025, Meta has used interactions with its AI features to personalize content and ads in most regions, and says it won't use conversations about sensitive topics such as health and religion to show ads.⁶

Google says Gemini chats aren't currently used to show you ads and that it will tell you if that changes.³

At OpenAI and Anthropic, deleted chats leave their systems within 30 days, with exceptions for legal and safety reasons.¹⁰⁷ Feedback is different: Anthropic keeps thumbs-up and thumbs-down reports, with the whole conversation, for up to 5 years.¹³

Training can't be reversed. Anthropic says that if you turn training off, your data stays in training runs already under way and in models already trained; it just won't be used in future runs.⁷

Most assistants let you download your history, but an export is a copy, not a move. ChatGPT and Claude offer self-service exports to individual users, and Microsoft offers Copilot history export through its privacy dashboard.¹⁹⁴ Anthropic notes that an export can't be imported into another personal Claude account.⁹

The law backs some of this. In the European Union, the General Data Protection Regulation (GDPR) gives a right to data portability in Article 20 for data you provided under consent or a contract.¹⁹ California's privacy law requires businesses to provide your data in a portable, machine-readable format where technically feasible, within 45 days of a verified request.²⁰

The Federal Trade Commission (FTC) has warned that quietly rewriting a privacy policy to start using customer data for AI training could be unfair or deceptive.²¹ It has also said AI companies are bound by the privacy promises they make, noting that past FTC orders required companies to delete models built from improperly obtained data.²²

In September 2025 the FTC ordered seven companies with consumer chatbots, including Alphabet, Meta, OpenAI, Snap and xAI, to explain how they use and share personal information from conversations.²³ It's a study, not an enforcement case.

In the EU, a proposal to amend the GDPR would spell out when companies may use personal data to develop AI under "legitimate interest"; as of October 2026 it was still being negotiated.²⁴

Our step-by-step guide, Lock down your AI assistants, shows where each of these switches lives. Settings and terms change, so recheck them whenever a provider emails you about an update.

• ThreatHomeBusiness

  AI chatbot data leaks and shadow AI

  Whatever you paste into an AI assistant can be stored, read by reviewers or used to train future models, depending on the app, the plan and your settings. In a 2025 survey of workers in 47 countries, 48% of those who use AI at work said they had uploaded sensitive company or customer information into public AI tools. The fix is mostly unglamorous: business accounts for work, the right settings at home, and a clear rule about what never goes in.

  Read
• ThreatHomeBusiness

  Prompt injection and AI agent hijacking

  When an AI assistant reads your email, files or web pages, it can mistake text written by a stranger for instructions from you. That is prompt injection, and when the assistant can also send, buy or delete things, an attacker can steer it. The UK's National Cyber Security Centre (NCSC) says it may never be fully fixed, so the safe move is to limit what your AI can reach and keep yourself in the approval loop.

  Read
• GuideHomeBusiness

  Lock down your AI assistants

  Each major AI assistant has a training setting, a memory setting, connected apps and a way to delete history, and the defaults favor the vendor. This guide walks through ChatGPT, Claude, Gemini, Microsoft Copilot and Meta AI one at a time, with notes for business accounts. Plan on about five minutes per assistant.

  Read
• GuideBusiness

  A one-page AI use rule for your team

  Your team is probably using AI already, so the useful question is which tools, with what data, and who checks the result. This guide helps you write those answers on one page, get everyone to sign it, and review it twice a year. When you need more, our full AI Acceptable Use Policy generator uses the same structure and is free with a free account.

  Read
• GuideBusiness

  Digital hygiene for your business: the 1-50 person baseline

  A small business's security baseline fits on one page: passkeys or multi-factor authentication (MFA) on every account, email and banking first; a shared password manager; automatic updates and endpoint protection; and backups you've actually restored from. Add a second-channel check before any payment change, a one-page AI rule, same-day offboarding and a short incident plan. Business email compromise alone drew $3.05 billion in reported losses to the FBI in 2025.

  Read
• ExplainerHomeBusiness

  Digital sovereignty, in plain English

  For governments, digital sovereignty means control over their own technology, data and online rules. For you, it means staying in charge of your own data, identity and choices: knowing who can read or train on your information, being able to leave a service with your data, holding the keys to your own accounts, and deciding what stays on your own devices.

  Read