Our AI assistants are trained on curated, lawful datasets that exclude copyrighted materials. The table below summarizes the types of data used in training, ensuring transparency and compliance with intellectual property laws:
| Data type | Description | Source | Exclusions |
|---|---|---|---|
| Proprietary QnA datasets | Question-and-answer pairs on compliance frameworks (e.g., ISO 27001, SOC 2, GDPR) | Created by Better ISMS based on consulting expertise, informed by general ISMS principles | No ISO NIST frameworks, or copyrighted content |
| Synthetic QnA pairs | Hypothetical compliance scenarios and questions | Generated by Better ISMS using original content | No copyrighted standards, regulations, or external texts |
| User-derived QnA insights | QnA pairs based on anonymized patterns from user conversations (e.g., common ISMS queries) | Insights from reviewing ISMS Copilot usage, stripped of user information | No user data or copyrighted materials |
| Public domain resources | Cybersecurity guidelines and expired copyright works | U.S. NIST publications (e.g., SP 800-53), public domain in the U.S. | No ISO standards or copyrighted content |
| Open-source content | Guides and reports under permissive licenses | Creative Commons (CC0, CC BY) materials from cybersecurity communities | No restricted TDM or copyrighted standards |
| Licensed content | Partners-provided original information security guidance | Generated by partners using orginal content | No ISO standards or copyrighted content |
Read mode: