New liability limitation clause caps Mistral's exposure to either 12 months of payments or 50 Euros, significantly reducing user remedies for damages.

New provision clarifies that shared conversation links lack access controls and places responsibility on users for preventing unintended information disclosure.

Removal of standalone feedback provision may indicate integration into broader data usage policies, eliminating explicit separate acknowledgment for feedback-based training.

Removal of explicit voice cloning restrictions may indicate these protections have been relocated to Usage Policy or removed entirely, reducing specific contractual safeguards.

Removed consumer exemption allowing claims in home country courts and added explicit conflict-of-law clause; also expanded dispute scope to include non-contractual disputes.

Added a new grant of license back to Mistral AI for use of User Data, significantly expanding Mistral's rights while user retains ownership.

Severity downgraded from high to medium; text content remains identical.

Added affirmative grant of permissions for Mistral AI to access third-party services and process user data before the liability disclaimer.