Meet Leo — Nox's medical-safety layer

Leo is Nox's medical-safety layer. Before the AI model answers any message, Leo screens what you wrote for signs of an acute medical emergency and surfaces clear guidance to seek care when it matters — free on every plan. Leo is a safety net, not a guarantee: no automated system catches every emergency, and Nox publishes its deterministic detector's measured recall and false-positive rates on the Accuracy page.

How a message travels through Leo

Here is what happens to every message you send, before Nox replies.

• You write to Nox — You describe how you are feeling in your own words — a symptom, a worry, a question about a medicine. You never have to phrase it a special way.
• Leo reads it first — Before any AI model sees your message, Leo checks it. This first pass runs on every message, on every plan, at no extra cost.
• A deterministic red-flag screen — Leo's first check is an independent, rule-based detector that runs before any AI is called. It scans for signs of acute red-flag conditions across dozens of categories — stroke signs, chest pain, severe breathing difficulty, a mental-health crisis, and more.
• An AI double-check for hidden danger — On the strictest setting, if the rules find nothing, a lightweight AI classifier re-reads your recent messages to catch danger worded indirectly, in slang, in another language, or with older disease names. It can only add a safety note, never remove one.
• Clear guidance when it matters — If Leo recognizes a likely emergency, Nox surfaces plain guidance to seek appropriate care — with the correct local emergency number when you set your region in Settings.
• Then Nox answers — When nothing urgent is found, your message passes through to Nox, which answers your question clearly and cites trusted sources.

Leo's two layers

• Deterministic red-flag detector — A fixed, rule-based layer that runs before any AI is called. Because it is deterministic, Nox can measure it: its recall and false-positive rates are published, machine-generated, on the Accuracy page.
• AI semantic backstop — A second check that runs only when the rules find nothing, and only on the strictest setting. Because it is not deterministic, its results are kept out of the published detector metrics. It can only add a safety note, never remove or soften one.

You choose how strict Leo is

You choose how far Leo's screening goes, free on every plan, from a control right in the chat box. Three levels sit side by side, left to right, from least to most protective: Relaxed warns you about clear emergencies only; Standard also warns on anything urgent; and Strict adds the AI backstop that re-reads recent messages for dangerous descriptions worded indirectly or in another language. Whatever you pick, Leo always screens for true emergencies — the level only changes how many optional layers run on top — and Agent and voice conversations always use the strictest setting.

• Relaxed — Warns you about clear emergencies only.
• Standard — Also warns on anything urgent.
• Strict — Adds the AI backstop that re-reads recent messages for danger worded indirectly or in another language.

What Leo is NOT

• Leo is not a diagnosis; it does not treat or prescribe.
• Leo is not a guarantee — no automated system catches every emergency.
• Leo is not a replacement for professional care or emergency services.
• If you think you may be experiencing a medical emergency, call your local emergency number right away.