Data and securityLimits and conditions
Data and security:limits and conditions.
Which data can be accepted, what agreed reuse may involve, and which records fall under other providers’ terms.
Data that cannot be accepted
- Identifiable or clinical data
- Data that identify or could re-identify individuals, clinical data, data whose rights are unclear, and data whose terms forbid external services are refused.
- No anonymization guarantee
- SciGround does not anonymize data or guarantee anonymity.
Reuse, only when agreed
- What may be used
- If your organization agrees to it, limited derived fields—such as workflow steps, method and checker versions, reason categories, and check status before and after a fix—may be used to build research tasks for evaluating and training research AI.
- What is excluded
- Free text, conversations, original files, measured values, study names, direct identifiers, and traceable IDs are excluded by default. There is no transfer to third parties, a parent company, or a successor by default, and no automatic model training.
- Stopping and deletion
- Stops and deletion requests are applied going forward, and their reach into tasks or copies already created is tracked; complete erasure of copies already provided is not guaranteed. Past uploads are never used retroactively.
Other providers and guarantees
- Infrastructure logs
- Cloud and sign-in providers keep their own management logs under their own terms. SciGround’s retention periods apply to the records SciGround collects, and are extended—for example, for a legal hold—only with explicit approval.
- Uploads outside the card
- Command-line hosts upload through the helper. If your AI has already read a local file, that happened outside SciGround.
- Verification is not a privacy guarantee
- Mathematical verification describes results and their conditions. It does not establish privacy, local-only processing, or data residency.