AIgree
← back

Mistral AI vs ChatGPT

Side-by-side comparison of the Terms of Service and Privacy Policy of Mistral AI and ChatGPT.

Mistral AI logo
★★★★☆
Mostly user-friendly

The documents include several consumer protections: users own outputs, can export/delete data, can object to training, have a 14-day withdrawal right, and can sue in their home-country courts. The main drawbacks are training use in some cases, long retention for certain records, auto-renewing subscriptions, and fairly broad moderation/termination powers.

Mistral AI’s terms are fairly user-oriented on ownership, withdrawal rights, and dispute venue, with some strong GDPR-style rights and export controls. At the same time, the service uses prompts, outputs, feedback, and usage data for product improvement and sometimes model training, retains some API data for 30 days and some records for years, and allows automated moderation and broad content controls.

Points of interest

  • positive ●●●●● terms
    Export and delete tools

    The service says you can export your data from your account and delete your account at any time, with a Help Center procedure for deleting data and digital assets. That makes it easier to leave and take your data with you.

  • negative ●●●●○ terms
    Auto-renewing subscriptions

    Paid subscriptions renew automatically until you cancel, so you need to actively manage billing to avoid unwanted charges. Failed payments can also lead to suspension or downgrade.

  • negative ●●●●○ privacy
    Training uses your content

    Mistral may use your input, output, and feedback to train its models in certain cases, including free subscriptions and feedback submissions. If you do not opt out where available, your prompts and outputs may help improve the model.

  • positive ●●●●○ terms
    You own outputs

    Mistral says you own your outputs, which is good if you want to reuse, publish, or commercialize what the model generates. It also says you retain ownership of your inputs.

  • positive ●●●●○ privacy
    Can object to training

    Mistral says you can object to the use of your input and output data for model training from account settings. That gives users a meaningful control over one of the most sensitive uses of their content.

  • positive ●●●●○ terms
    Home-country courts apply

    For EEA consumers, you can bring claims in the courts of your country of residence, and your home-country substantive law applies. That is more user-friendly than forcing a distant forum or foreign law.

  • negative ●●●○○ privacy
    Long retention periods

    Le Chat conversations stay until you delete them, API inputs and outputs may be kept for 30 days for abuse monitoring, and some records are retained much longer, including invoices for 10 years. That means a lot of data can persist after use.

  • negative ●●●○○ terms
    Broad moderation controls

    Mistral reserves the right to monitor usage with automated tools and to remove or restrict data that violates its rules or creates risk. This can affect availability of your content or account if the service flags it.

  • positive ●●●○○ terms
    Withdrawal right offered

    You have a 14-day withdrawal right after subscribing, and Mistral says it will refund payments within 14 days after notice. This is a standard but important consumer protection for paid plans.

  • neutral ●●○○○ terms
    Conversation links are public

    If you share a conversation link, anyone with the link can view it, and Mistral says it does not control who accesses it. This is not a hidden clause, but it is easy to overlook and has obvious privacy implications.

  • neutral ●●○○○ privacy
    No US data sale or ad sharing

    For certain U.S. residents, Mistral says it has not sold, shared, or used personal data for targeted advertising in the previous 12 months. That is a reassuring privacy statement, though it is limited to a defined period and jurisdiction.

Documents

ChatGPT logo
ChatGPT
AI
★★★★☆
mostly user-friendly with caveats

The documents offer strong transparency, deletion/export rights, and local-court dispute resolution, which are helpful for users. The main concerns are broad content-processing rights, some retention beyond deletion, and extensive sharing for service, security, and legal purposes.

ChatGPT’s consumer terms are relatively user-protective on dispute rights and deletion controls, with no mandatory arbitration in the EEA/UK version and explicit access to local courts. The main tradeoff is broad permission to use user content to operate, improve, and secure the service, plus model-training use unless you opt out. Privacy disclosures are detailed, with export/deletion tools and statutory rights, but they also describe sharing with vendors, affiliates, and authorities, and retention beyond deletion for security, legal, and fraud reasons.

Points of interest

  • negative ●●●●○ terms
    Broad content usage rights

    OpenAI can use your content worldwide to provide, maintain, develop, improve, and secure the services, plus enforce terms. For users, this is a wide license covering very broad operational uses.

  • positive ●●●●○ terms
    Local courts available

    If a dispute cannot be resolved informally, the terms let either side go to local courts instead of forcing arbitration. That preserves a user’s ability to sue in court.

  • positive ●●●●○ privacy
    Delete and export tools

    The privacy policy says you can delete chats or your account and export your history and data. That makes it easier to leave or clean up your data than with many services.

  • positive ●●●●○ privacy
    Statutory privacy rights listed

    The policy expressly lists access, deletion, correction, portability, restriction, consent-withdrawal, and complaint rights. Users also get instructions for submitting requests.

  • negative ●●●○○ privacy
    Model training uses content

    The privacy policy says some content may be used to improve services and train models unless you opt out. That means chats and uploads may contribute to model development by default in some cases.

  • negative ●●●○○ privacy
    Broad data sharing

    OpenAI shares personal data with vendors, affiliates, business account administrators, and sometimes government authorities or other third parties. Users should expect their data to leave the service in multiple scenarios.

  • negative ●●●○○ privacy
    Retention beyond deletion

    Deleted data may still be kept for up to 30 days, and longer for legal, security, fraud, or accounting reasons. So deletion is not always immediate or absolute.

  • positive ●●●○○ terms
    Training opt-out available

    OpenAI says it may use your content to train models, but you can opt out in account settings. That gives users some control over whether their chats help improve the system.

  • negative ●●○○○ terms
    Auto-renewing subscriptions

    Paid subscriptions renew automatically until canceled, so users need to manage cancellation themselves to avoid further charges. The terms do provide a 14-day cooling-off period for EEA consumers.

  • positive ●●○○○ privacy
    Temporary chats auto-delete

    Temporary Chats are described as being automatically deleted within 30 days, and Atlas incognito browsing history is not saved after a session ends. This can reduce long-term retention for some activity.

Documents

Comparison is based on each service's published Terms of Service and Privacy Policy. Read the source documents linked above before relying on any specific clause.