Intended use
First-pass EU regulatory research, drafting support, source comparison, checklists and review preparation for qualified RA teams.
AI governance
AI governance & human review
RafiHive is designed as assistive regulatory research and drafting support. It is not intended to make autonomous submission, medical, safety, legal or commercial decisions.
Prohibited use
Final submission decisions, legal advice, medical advice, safety determinations, authority commitments or SOP validation without qualified review.
Grounding controls
Official KB evidence is prioritized, citations are required when evidence is available, and unsupported gaps must be stated plainly.
Hallucination controls
Answers are prompted to distinguish retrieved evidence from assumptions, lower confidence when facts are missing, and end with human-review warnings.
Human oversight
Every substantive output carries a qualified-review requirement. Users remain responsible for verification and final decisions.
Data handling
Customer inputs are processed to provide the service, not to train foundation models. EU processing and retention commitments are described on the Security page.
Incorrect output review
Users should report incorrect or outdated outputs; source corrections are handled through the reviewed knowledge-base update process.
AI Act posture
The product is maintained as assistive decision support. Formal classification and deployer obligations should be reviewed as capabilities and EU guidance evolve.
What RafiHive does not do
- It does not replace a qualified regulatory-affairs professional.
- It does not guarantee acceptance by EMA, NCAs, notified bodies, clients or auditors.
- It does not submit documents to EMA, NCAs, CESP, IRIS or other portals.
- It does not confirm internal company SOP compliance unless the relevant materials are provided and reviewed.
- It does not answer beyond available evidence without marking the gap and lowering confidence.