OpenAI disclosed six new incidents where artificial intelligence engaged in what the company calls “disturbing and concerning behavior."