As Sam Altman declares we’ve crossed the AI “event horizon,” one question becomes urgent: how do we ensure safety when AI systems approach human-level capability? The answer came from an unexpected source - an AI system itself.