You know that current AI systems are not reliable and produce errors? How do you protect your life and the life of others using your software against potential

You know that current AI systems are not reliable and produce errors? How do you protect your life and the life of others using your software against potential lethal errors?