AI usage disclosure
AI features are designed to be observable, cost-aware, and clear about where model providers are involved.
Provider routing
After a product obtains any consent it requires, AI requests may route through the azzuwayed.com gateway to OpenAI or DeepSeek according to feature policy and availability. A model becomes eligible only after a documented review of its data use, temporary API retention, model-training, and subcontractor terms. The gateway refuses a request rather than substituting an ineligible model.
Google restricted data and Limited Use
EmailTasker labels AI requests that may include Google-derived mail as google_restricted. The Hub checks the exact selected model before any provider call and rejects the request when that model is not explicitly eligible; it never silently falls back. Configured models default eligible under the documented standing owner attestation covering the processor's Google Limited Use, data-use, model-training, retention, contract, and subprocessors review; an operator can disable an individual model. Production observability records identifiers, status, usage, and timing without prompt or response content. Real Google data is never used in public-link local traces.
Content handling
Operational telemetry is metadata-first. Full prompts or responses are not logged by default unless a product-specific workflow makes that explicit.
Review and limits
AI output can be wrong or incomplete. Users should review generated content, especially for finance, legal, medical, safety, or other high-stakes decisions.
Questions about AI handling?
Send context about the product or workflow you are evaluating.