Why does AI screen identification mask values before calling the model?

Before any model call, TermIQ masks DYNAMIC regions (values that change across snapshots): letters become X, digits become 9. Masked copies go to sanitizedData; the original snapshot data is never overwritten. Labels and static structure stay readable so the model can name the screen without seeing session-specific data. Prefer multiple snapshots per screen so dynamism — and therefore masking — is accurate.