I heard this NPR storythis news story about Professor Ethan Mollick, who addressed this problem by requiring students to work with ChatGPT to write their essays. This involved improving its drafts, looking up sources, stuff like that.
Other educators are doing so as well, per this ABC news story.
Can you do something similar? Give students the option to turn in ChatGPT's response, but they have to say that is was generated this way, and either explain why it's wrong, or prove somehow that it's correct (like with a different method of solving).