Privacy grade breakdown
Weighted from the six policy factors below. The formula lives in the methodology.
Policy details
What the company’s own documents state, with the nuance kept.
Default model training
Cursor does not use Inputs or Suggestions to train models by default, and does not let third parties train on them, unless content is flagged for security review, you submit Feedback, or you explicitly agree to training. Privacy Mode (when enabled) also enforces zero data retention with model providers. Turning Privacy Mode off allows Cursor to store and train on codebase/prompt data.
Opt-out mechanism
Keep Privacy Mode enabled in Cursor settings (and online account settings) so training stays off and ZDR applies. Only disable Privacy Mode if you accept training and non-ZDR retention.
Human review of chats
Staff do not read Google-connected Grok Bot data unless you ask for support, for security investigations, or to comply with law. Security-flagged Inputs/Suggestions may be analyzed to improve abuse detection.
Third-party sharing
Shared with vendors (hosting, cloud, model providers, analytics, support, safety, payments). Subprocessors for commercial processor use are listed at trust.cursor.com/subprocessors. Model providers under Privacy Mode operate under ZDR agreements except designated non-ZDR models.
Data retention and deletion
Retained as needed to operate the service, legal compliance, safety, and dispute resolution. Privacy Mode uses ZDR with providers; temporary encrypted file cache exists for latency and is not used as training data when Privacy Mode is on. Some temporary interactions may be stored briefly for safety/monitoring.
Request access/deletion via hi@cursor.com or account controls. Data is deleted, erased, de-identified, or anonymized when no longer needed.