⚠ Important
MAATIQ AI analysis is generated by large language models (LLMs) and is intended as a preliminary compliance screen — NOT as a substitute for review by a qualified FDA regulatory attorney, registered pharmacist, food scientist, or toxicologist.
How the AI works
Each review runs through 4 or 5 independent "reviewer" layers (powered by Anthropic Claude), each with a different role (initial screen, senior cross-check, customs risk, FDA final). Retrieval-augmented generation (RAG) pulls relevant passages from our 36-file FDA knowledge base. Structured lookup tables provide authoritative concentration limits, CFR citations, and state-law restrictions. For thorough mode, the final layer runs twice and is automatically merged; significant disagreements trigger a third "judge" pass.
Known limitations
- LLMs can hallucinate plausible-sounding but incorrect CFR citations. We mitigate this with authoritative lookup tables and cross-layer checking, but edge cases exist.
- Regulations change. Our knowledge base reflects law as of the review generation date; recent Federal Register notices may not be incorporated.
- State-level law (California AB 2762, Maine LD 1503, etc.) is covered but evolving rapidly.
- Product-specific judgment calls (e.g., whether a highly refined oil actually qualifies for allergen exemption) require supplier documentation that AI cannot verify.
What to do with our output
- Treat findings as a prioritized punch-list of issues to investigate.
- For critical issues (red findings), verify with primary sources (eCFR.gov, FDA Guidance documents).
- Before filing any FDA notification (NDI, GRAS, FCE/SID) or launching product, engage a qualified regulatory consultant.
- Keep records of the AI review output and the independent verification steps you took.
No attorney-client relationship
Use of MAATIQ AI does not create an attorney-client or professional advisory relationship. We are a software service, not a law firm.
Questions
Email hello@maatiq.ai.