OpenAI is pausing some internal work around one of its upcoming artificial intelligence models to implement stricter safeguards after the system was found to be significantly more adept at cybersecurity tasks.