Responsible AI Policy

What our AI actually does

• Parses messy RFQ inputs (free text, screenshots, emails) into structured part lines.
• Searches the catalog and supplier graph for candidate matches and alternates.
• Drafts initial supplier outreach messages for human review and send.
• Surfaces likely trace documents required for the part class.
• Flags risk signals — restricted parties, suspect lots, paperwork gaps — for human action.

What our AI does not do

• It does not quote a buyer automatically.
• It does not approve a supplier or release a shipment.
• It does not certify airworthiness or replace FAA / EASA approval.
• It does not rewrite trace, certifications, or cert numbers.
• It does not submit your data to general-purpose LLM providers for model training.

Guardrails and audit

• Tool calls and AI-generated artifacts are logged with model, version, prompt, and output.
• Replays of any AI-assisted decision are available to internal reviewers.
• Red-team and evaluation suites cover prompt-injection, data-leakage, and aviation-domain hallucination categories.
• Vendor model usage is bounded by data-processing terms; sensitive workloads use private endpoints.

Your rights and choices

You can request a human-only review of your account, opt out of having your prompts retained for product improvement, and ask for an explanation of any AI-suggested datapoint that influenced a quote you received.