Nearly 10% of people ask AI chatbots for explicit content. Will it lead LLMs astray?

Beyond programming tips and writing help, a million conversations reflect people’s desires for other kinds of ‘usafe’ information. Here’s what researchers are doing about it.