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
★★★★☆
Fairly user-friendly

The documents offer strong transparency, EU consumer protections, account-level deletion/export controls, and no mandatory arbitration. However, prompt/output data may be used for training by default in some plans, retention can be lengthy for some records, and subscriptions auto-renew with generally non-refundable payments.

Mistral AI’s EU consumer terms are relatively transparent and include useful user controls like export, deletion, and an opt-out for model training. The main tradeoffs are that some prompts/outputs may be used for model training unless you opt out, conversations can persist until deletion, subscriptions auto-renew, and liability is limited. The service does preserve court access for EEA users and does not impose mandatory arbitration.

Points of interest

  • negative ●●●●○ terms
    Training on user content

    Mistral may use your prompts and outputs to train models on free subscriptions and some Le Chat paid plans unless you opt out. That means your AI interactions may help improve the model by default in those cases.

  • positive ●●●●○ terms
    No mandatory arbitration

    EEA consumers can bring claims in their home-country courts or in Paris, and mediation is voluntary rather than forced. This preserves the practical right to sue in court.

  • positive ●●●●○ privacy
    Training opt-out control

    You can object to use of your input/output for model training directly in account settings. This gives users a meaningful privacy control without needing to contact support first.

  • positive ●●●●○ privacy
    Export and deletion tools

    Mistral provides self-serve controls to export data and delete your account, plus a switching process to move to another provider. This improves portability and makes exit less burdensome.

  • negative ●●●○○ privacy
    Conversations kept until deleted

    Le Chat inputs and outputs may remain stored until you delete the conversation or your account. If you do not actively clean up chats, they can persist indefinitely.

  • negative ●●●○○ privacy
    Long record retention

    Some account-related records are retained for years, including invoices for 10 years and certain identity/support records for 5 years. Even after account deletion, not all data disappears immediately.

  • negative ●●●○○ terms
    Broad moderation powers

    Mistral may automatically monitor use, review content, and remove or restrict data or suspend accounts for breaches, legal compliance, or risk. This can be important if you rely heavily on the service.

  • positive ●●●○○ privacy
    No sale or targeted ads

    For covered U.S. users, Mistral says it did not sell, share, or use personal data for targeted advertising in the past 12 months. This is a favorable signal against ad-tech style exploitation.

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

    Paid plans renew automatically until canceled, and cancellation only stops the next renewal rather than producing a prorated refund. Users need to cancel in advance to avoid another charge.

  • negative ●●○○○ terms
    Mostly non-refundable payments

    Payments are generally non-refundable, although EEA consumers do get a 14-day withdrawal right. After that window, canceling usually just preserves access through the current term.

  • negative ●●○○○ terms
    Liability is limited

    The company excludes liability for unforeseeable losses, losses not caused by its breach, and events beyond its control, while preserving non-excludable consumer rights. That may narrow recovery for some service problems.

  • negative ●●○○○ terms
    Terms can change

    Mistral may update the terms and privacy policy over time. The terms promise 30 days' notice for material changes, which is better than silent updates but still allows unilateral changes.

Documents

ChatGPT logo
ChatGPT
AI
★★★★☆
Fairly user-friendly

The documents provide strong consumer rights for covered users, including local-court dispute resolution, preserved statutory rights, meaningful privacy controls, deletion/export tools, and a training opt-out. Main concerns are broad data collection, model-training use by default, auto-renewal, admin access for work accounts, and retention exceptions.

ChatGPT’s EEA/UK/Switzerland terms are relatively consumer-protective compared with many online services: users keep input ownership and generally own output, can go to local courts, and get clear account controls such as deletion, export, training opt-out, and temporary chats. At the same time, OpenAI collects broad usage and content data, may use content to improve models unless you opt out, auto-renews paid plans, and reserves rights to suspend accounts and retain some data for safety, legal, and business reasons.

Points of interest

  • positive ●●●●● privacy
    Strong deletion and export controls

    Users can delete chats, delete accounts, export data, use temporary chats, and manage memory settings. Deleted personal data is generally removed within 30 days unless an exception applies.

  • negative ●●●●○ terms
    Content used to improve models

    OpenAI can use your prompts, uploads, and outputs to develop and improve services, including model training, unless you opt out. This creates a meaningful privacy tradeoff for sensitive chats.

  • positive ●●●●○ terms
    No forced arbitration

    Disputes can go to your local courts, which is more user-friendly than mandatory arbitration or class-action waivers. EEA users also get access to an EU dispute resolution platform.

  • positive ●●●●○ terms
    Statutory rights preserved

    The terms explicitly say consumer rights under applicable law are not waived. This helps protect users against overbroad contract terms.

  • positive ●●●●○ terms
    You own input and output

    Users retain ownership of their input and, where legally permitted, own the output assigned by OpenAI. That is stronger than services that claim broad ownership over user-generated results.

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

    You can turn off use of your content for model training in account settings. This is a significant privacy control, even though it is not the default.

  • negative ●●●○○ privacy
    Broad data collection

    The privacy policy allows collection of account data, payment data, prompts, files, contacts, device, usage, cookies, location, and some outside-source data. Users should assume extensive telemetry and content processing.

  • negative ●●●○○ privacy
    Retention exceptions apply

    Although deletion tools exist, OpenAI may keep data longer for legal, security, fraud, abuse, and accounting reasons, and de-identified training data may remain disassociated from your account. So deletion is not always immediate or absolute.

  • negative ●●●○○ privacy
    Work admins can access content

    If you use an employer or business-linked account, administrators may control the account and access your content, and your organization may learn you have the account. That reduces privacy in workplace contexts.

  • positive ●●●○○ terms
    Clear notice of term changes

    Material adverse changes to the terms require at least 30 days' advance notice. This is better than immediate unilateral changes without warning.

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

    Paid plans renew automatically until canceled. After the 14-day cooling-off period, cancellation usually stops future charges but does not refund the remaining billing period.

  • neutral ●●○○○ privacy
    International data transfers

    Personal data may be processed outside Europe, including in the United States, using adequacy decisions or standard contractual clauses. This is common, but some users may prefer local-only processing.

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.