The company is expanding its efforts to protect ChatGPT users in cases where conversations may turn to self-harm.