A new survey reveals that although security leaders see AI as superior at identifying threats, the widespread adoption of large language models is not likely on the horizon.