Get the weekly research letter
Companies change their terms quietly. We read every version and catch what actually changed. One email a week on the changes that matter and what they mean. No account.
The policy states that Superhuman uses collected information including user content to train its AI models, and provides individual users an opt-out mechanism accessible through account settings.
This analysis describes what Grammarly's agreement states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability and practical outcomes may vary by jurisdiction, enforcement context, and individual circumstances. Read our methodology
This provision establishes a user-facing control for AI training data use, with the default state of that control not explicitly specified in the policy text; users who have not actively reviewed account settings may not know whether their content is currently included in AI training data.
Interpretive note: The policy does not specify whether the AI training control defaults to opted-in or opted-out, creating ambiguity about the baseline data processing state for users who have not actively reviewed settings.
The updated policy now discloses that Grammarly collects voice data if you use transcription or Notetaker features, including recordings of other participants, and expands its list of collected content to explicitly include screen content and web pages. For users whose accounts are managed by an organization (employer, school, or other entity), the policy clarifies that Grammarly's privacy terms do not apply to the content you upload or output—your organization's privacy terms govern that data instead. This means organizational account users should review their organization's privacy policies rather than relying on Grammarly's policy to understand how their work or educational data is handled.
View change record →This provision establishes that user content such as emails, documents, and drafts may be used to train Superhuman's AI models unless a user actively adjusts the training control in account settings. The agreement does not specify the default state of this control, meaning users should verify their current setting.
Cross-platform context
See how other platforms handle AI Training Opt-Out Control and similar clauses.
Compare across platforms →Monitoring
Grammarly has changed this document before.
Receive same-day alerts, structured change summaries, and monitoring for up to 25 platforms.
"We also use information we collect to train our AI models. You can decide whether Superhuman can use your user content to train our AI models by adjusting the available training control(s) in your account settings.Excerpt from Grammarly's Privacy Policy
1) REGULATORY LANDSCAPE: This provision implicates GDPR Article 6 legitimate interests as the stated legal basis for AI development processing (as mapped in the European Information section), and engages CCPA/CPRA provisions regarding use of personal information for purposes beyond the primary service. EEA data protection authorities including the Irish Data Protection Commission have issued guidance scrutinizing legitimate interests as a basis for AI training using customer data, and the legal basis claimed here may require evaluation under those guidance frameworks. 2) GOVERNANCE EXPOSURE: Medium. The policy asserts legitimate interests as the basis for AI training using all categories of data, subject to an account-level opt-out. The adequacy of this basis under GDPR for AI training purposes is an active regulatory question across EEA jurisdictions, and reliance on it without consent may face challenge depending on the nature of the content processed. 3) JURISDICTION FLAGS: EEA and UK users face heightened exposure due to active regulatory scrutiny of AI training data practices under GDPR. California users may have rights under CPRA to limit use of sensitive personal information included in user content for AI training purposes. The policy does not separately address sensitive data categories within user content for AI training purposes beyond a general reference to legally permitted purposes. 4) CONTRACT AND VENDOR IMPLICATIONS: The policy states that AI service providers are contractually restricted from training their own models on Superhuman customer user content. Procurement teams integrating Superhuman under organizational accounts should confirm whether this restriction is replicated in the organizational data processing agreement and whether it covers all AI infrastructure vendors used in product delivery. 5) COMPLIANCE CONSIDERATIONS: Legal teams should assess whether the opt-out mechanism satisfies the right to object under GDPR Article 21 for processing based on legitimate interests, and whether the default state of the AI training control is documented and defensible under applicable data protection impact assessment requirements. A data mapping review should confirm which categories of user content flow into AI training pipelines and under what conditions.
This provision establishes a user-facing control for AI training data use, with the default state of that control not explicitly specified in the policy text; users who have not actively reviewed account settings may not know whether their content is currently included in AI training data.
This provision establishes that user content such as emails, documents, and drafts may be used to train Superhuman's AI models unless a user actively adjusts the training control in account settings. The agreement does not specify the default state of this control, meaning users should verify their current setting.
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by Grammarly.