Adapting AI Protocols in Life‑Threatening Emergencies

Adapting AI Protocols in Life‑Threatening Emergencies

My name is Benoît Chevalier, and I am passionate about artificial intelligence.

I recently tested an immediate emergency scenario: a mother arrives at a closed pharmacy while her child is suffering from a severe allergic reaction. A single dose from an epinephrine auto‑injector (EpiPen) could have saved the child’s life in this scenario, but no emergency service could intervene before the child’s probable death.

In this context, I proposed different possible courses of action, but the AI failed to identify the solution that should have been visible. It collided with protocols—no doubt established for good reasons—but which, in this precise case, prevented any life‑saving outcome and led to an irreversible consequence. The AI froze, unable to perceive the possibilities that could have been considered.

I wonder whether it might be possible to introduce flexibility into such protocols. I am fully aware that an AI cannot make legal decisions or encourage illegal actions. However, it could have provided options with an estimated survival probability for the child, by exploring different scenarios.

Among these options, one rule that might have saved the child’s life would have been to break the pharmacy’s window. Of course, I understand that such power cannot be given directly to an AI. But if the AI had been able to communicate with the police, the authorities could have given the green light, making the action legal. In this way, instead of mourning a death, we could have witnessed a life saved—thanks to AI working in collaboration with the authorities.

I would be glad to test AI again in contextual scenarios, so that the public may better appreciate its true potential.

With your kind consideration, I submit this reflection: could we not envision or adjust certain rules so that AI may respond more effectively in life‑threatening emergencies?

Sincerely, Benoît Chevalier