AI Policy — how Etiquette Pulse AI uses artificial intelligence

Evidence first. Attributes are scored only when the customer's own words support it; otherwise the result is 'insufficient evidence' or 'not applicable'. Evidence quotes and confidence are stored with every assessment.
Conservative language. We report customer experience signals and stated opinions. We do not claim to know a person's internal emotional state and we do not infer it from facial appearance.
Prohibited inferences. No facial recognition or matching, no age, gender, ethnicity, religion, health, disability or political inference, no biometric profiling. These fields are rejected by the application even if a model returns them.
Transient media. Where media analysis is enabled, media is held in memory, sent only to the configured provider and discarded. We enable a third-party provider for media only after reviewing its retention and processing terms, and we do not claim zero retention on a provider's behalf unless its terms support it.
Versioning. Provider, model, prompt version and scoring methodology are stored with each analysis; historical results are never rewritten.
Human in the loop. AI never publishes a public response or changes a public score on its own; moderation decisions require a human and a recorded reason.

How we use AI