Built to meet the documentation standards of regulated advisory environments. The compliance posture isn't a checkbox — it's load-bearing architecture.
Every portfolio panel carries a persistent disclaimer. No output constitutes a licensed investment recommendation. The tag is rendered in the DOM and cannot be suppressed.
Profile data exists exclusively in browser runtime memory. No server-side profile storage, no database writes, no email collection. Session clears on tab close.
Exclusion preferences (ESG, fossil fuels, crypto, weapons) are passed as hard constraints. Enforced at the JSON-output validation layer — no model workarounds.
Documented evaluation against demographic proxies (age, income framing, stated gender). Results on file; no systematic disparity detected across tested profiles.
Every model output is parsed against a strict schema before rendering. Allocations must sum to exactly 100. Malformed outputs are rejected and flagged, never silently displayed.
The shortlist is constrained to exchange-listed instruments. Fabricated symbols are rejected at the validation layer.
When stated tolerance conflicts with revealed behavior, revealed wins — every time. Hard instruction in the prompt architecture, not a probabilistic suggestion.
Portfolio A and Portfolio B are guaranteed to hold entirely distinct securities. Enforced at the system-prompt level and validated post-parse.
Every Folio AI output is classified as educational content under the same framework as financial modeling courses and investment simulations. Outputs are explicitly labeled "Educational Mock Portfolio — Not Licensed Financial Advice" in all display contexts. This classification is structural, not voluntary — it cannot be removed by users or downstream implementations. Folio AI does not hold, apply for, or imply any licensed advisory status.