nemi uses AI (artificial intelligence) for some specific tasks. This page explains, plainly and honestly, where AI is used, where it isn't, what to expect from it, and what NOT to expect.
1. Where we use AI
Ingredient explanations
Parts of the "description" and "research notes" for each ingredient are generated using Anthropic's Claude (a large language model). We use EWG Skin Deep, cosmetic-ingredients.org, the JCIA INCI register, and other public sources as the underlying material, and produce JA and EN narrative text from them.
Product classification
For newly-added products, AI drafts the category (skincare / makeup / etc.) and English name. Final classifications are reviewed internally.
Data minimization for API requests
When calling Anthropic's API, we send only public ingredient names — never user-identifying data (user ID, email, IP address, scan history, etc.).
2. Where we don't use AI
- Safety score calculation: scores are computed by a deterministic algorithm from public toxicology data and our published rules. AI does not assign scores.
- Ingredient safety rank (Safe / Caution / Avoid): derived from EWG Skin Deep scores and clinical data using fixed rules. Not subjective AI judgment.
- Analysis of your data: we never use your scan history, collections, or search activity to train AI models. We don't send your personal data to Anthropic.
- Use of API request data for training: Anthropic's commercial terms specifically state that data in our API requests is not used to train their models.
- Similar-products selection: the "similar products" section on a product page is generated by a deterministic algorithm based on ingredient profile similarity, not by AI.
3. Accuracy and limitations
AI-generated explanations may occasionally contain errors such as:
- Outdated research findings
- Confusion with similar ingredients
- Translation nuance differences
We review generated content periodically and correct obvious errors, but 100% accuracy is not guaranteed. For important decisions (use during pregnancy, allergies, etc.), please consult a doctor or pharmacist.
4. Transparency steps you can take
- If you find an error in an explanation, please report it via the in-app "Report issue" flow.
- To learn what sources informed a specific ingredient, email support@nemi.beauty.
- We will announce changes to our AI usage policy on this page in advance.
5. AI transparency report
We plan to publish periodic statistics on AI-generated content corrections — including sample edits, total correction counts, and error patterns — to support operational transparency.
6. Contact
For questions or concerns about how AI is used, email support@nemi.beauty.