What Happens When Leo Detects an Emergency?

Leo screens each message for signs of acute medical red flags before Nox’s AI-generated response appears. When it detects a likely emergency, Nox shows clear emergency or urgent-care guidance first, including a local emergency number when you have set your region.

That order matters. In a potentially time-sensitive situation, the priority is not a long explanation or a back-and-forth conversation. It is making sure the person sees guidance to seek appropriate care without waiting for an AI answer.

How Leo screens a health conversation

Leo is Nox’s user-facing medical-safety system. It looks for signs of acute red-flag situations across a wide range of categories, including stroke signs, chest pain and cardiac symptoms, severe breathing difficulty, severe allergic reactions, serious bleeding or trauma, overdose or toxicology concerns, pregnancy warning signs, concerning symptoms in children, environmental emergencies, and mental-health crises.

The first screen is deterministic: it uses fixed rules to check each message before any AI model is called. This design is intentional. A deterministic system produces the same result when it sees the same relevant pattern, which makes its performance measurable and reviewable.

When one of those rules fires, Nox presents an emergency or urgent-care banner before AI-generated content. Leo is not making a diagnosis or deciding exactly what is happening. It is recognizing that the description may involve a situation where prompt professional help is more important than an ordinary informational answer.

This is the core idea behind why health AI should escalate instead of diagnose: in a possible emergency, a safe system should direct attention toward getting care rather than attempting to settle uncertainty through conversation.

What you see when Leo detects a likely emergency

When Leo recognizes a likely emergency, Nox surfaces clear guidance to seek appropriate care ahead of any AI reply. If your region is set in Settings, the guidance includes the relevant local emergency number.

If Nox does not know your region, it falls back to universal emergency numbers: 911, 999, and 112. The goal is that emergency guidance is not left blank simply because a location has not been configured.

The appropriate next step depends on the guidance shown and the circumstances. For an immediate emergency, contact local emergency services. If someone may be in immediate danger and you can safely help, contact local emergency services on their behalf.

Leo’s guidance is designed to interrupt the usual flow of a health chat for a reason. In a possible acute emergency, an answer that is detailed, reassuring, or speculative can distract from the more important action: seeking timely in-person help.

Why the warning comes before the answer

Health conversations can be ambiguous. Someone may describe symptoms directly, use casual language, refer to an old name for a condition, or mention a concern indirectly. In lower-stakes situations, clarifying questions can be useful. In a potential emergency, delaying the safety message can create an avoidable risk.

That is why Leo’s deterministic screening layer runs before the AI response. The emergency or urgent-care banner is not dependent on the AI deciding whether to mention it after generating an answer.

This separation also helps define Leo’s role clearly. It is a safety net, not a medical evaluator and not a replacement for emergency services or a clinician. A red-flag warning means the conversation contains language that deserves urgent attention—not that Nox has confirmed a condition.

For a closer look at the system’s design, read how Leo works. The important practical point is simple: if Leo interrupts with emergency guidance, treat that guidance seriously and prioritize getting appropriate care.

What if an emergency is described indirectly?

Fixed rules are useful, but people do not always describe urgent problems with predictable wording. A person may use slang, write in another language, use older medical terminology, or describe a dangerous situation indirectly.

When Leo’s deterministic layer does not find a match, a second check can run in the background. This lightweight AI classifier rereads recent messages to look for descriptions that may still indicate a likely emergency.

If that backstop identifies a likely emergency, Nox shows the same seek-care guidance used for a deterministic pattern match. Importantly, the backstop can add a safety note, but it cannot remove or soften one that the deterministic layer has already triggered.

That one-way design is an important safety principle. A later system should not be able to overrule an earlier emergency warning and make the response less cautious. Rule-based safety systems vs. AI safety models explains why these approaches can serve different roles in a layered safety process.

Can you choose how sensitive Leo is?

Leo offers three screening levels: Relaxed, Standard, and Strict. You can select the level from a control in the chat box, and the setting is available on every plan.

Relaxed warns about clear emergencies only. Standard also warns about situations that may be urgent. Strict adds the AI backstop that rereads recent messages for danger expressed indirectly or in another language.

No matter which level you choose, Leo always screens for true emergencies. The setting changes how many optional protective layers are active beyond that baseline. Agent and voice conversations always use the strictest setting.

A stricter setting may surface more warnings, including some that turn out not to be emergencies. That tradeoff is part of cautious screening: a system designed to catch concerning language should not pretend it can perfectly distinguish every situation from a chat message alone.

What Leo cannot guarantee

Leo does not catch every emergency. No automated system can reliably identify every urgent situation, especially when symptoms are incomplete, changing, understated, or not mentioned in the conversation.

That limitation is not a reason to ignore symptoms that concern you. If you think you or someone else may be experiencing an emergency, contact local emergency services rather than relying on an app or waiting for a chat response. For concerning or persistent symptoms that are not an immediate emergency, contact a qualified clinician.

Nox publishes high-level measurements of the deterministic detector’s recall and false-positive rate on its Trust & Transparency page. Those metrics apply to the deterministic detector and are kept separate from the non-deterministic AI backstop.

Transparency does not make an automated system perfect. It does give people a clearer basis for understanding what the safety layer is designed to do, how it is assessed, and where its limits remain.

Common questions

Does Leo call emergency services for me?

Leo surfaces guidance to seek appropriate care and can show an emergency number based on your region setting. The product documentation does not state that Leo contacts emergency services on a user’s behalf.

Does a Leo warning mean I have a specific condition?

No. Leo does not diagnose. It identifies language that may signal an acute red-flag situation and prompts you to seek appropriate care.

Will Leo always catch an emergency?

No. Leo is a safety net, not a guarantee. If you believe there may be an emergency, contact local emergency services even if no warning appears.

Why might Leo warn me when I am unsure it is serious?

Leo is designed to prioritize safety when a message could describe an emergency or urgent situation. A warning is a prompt to take the possibility seriously and seek professional help when appropriate.