AI Ethics

Where AI is used in the product

• Session setup: generating or refining scenario context, airport/procedure context, and pilot-facing practice setup before a session starts.
• RT Lab dialogue: running the simulated ATC voice/text interaction during practice.
• Debriefs: producing transcripts, competency scores, and grounded observations after a session for review and export.

Where AI is not the authority

• Not operational ATC: InnoViate must not be used for live flight operations, dispatch, separation, or real-world ATC decision-making.
• Not a navigation source: InnoViate is not an approved EFB, charting product, dispatch tool, or operational navigation source.
• Not instructor sign-off: for organisation workflows, formal outcome and sign-off decisions remain with the instructor or admin, not the model.
• Not a guarantee of correctness: model output can be incomplete, wrong, or poorly timed. Users should treat it as a training aid, not as a sole source of truth.

Safety and grounding approach

• Standards-first: where the product evaluates phraseology or readbacks, we aim to tie outputs back to defined doctrine or training references instead of letting the model improvise unsupported authority.
• Fail closed where possible: if the system cannot confidently ground an answer, the preferred behavior is to ask for clarification, defer, or stay narrow rather than invent details.
• Deliberate launch scope: current realism data and procedure coverage are FAA-first and United States-focused at launch. We do not market that as global coverage.
• Bounded training realism: some scenario elements may be simplified or synthetic to support training objectives. Where realism is bounded, the product should not be treated as authoritative operational guidance.

Human accountability

We review failures, refine prompts and controls, and adjust system boundaries when behavior is unsafe, misleading, or ungrounded. If you see a problematic output, contact hello@innoviate.ai.