Linus’ thread: (CW: bigotry and racism in the comments) https://social.kernel.org/notice/AWSXomDbvdxKgOxVAm (you need to scroll down, i can’t seem to link to the comment in the screenshot)
Given the number of bots on the internet trying to crack captchas, this is already happening. I don’t think captchas are being used for AI training that much, since hCaptcha uses AI-generated images with prompts like “Select the images with a hamster eating a watermelon” for its tests. All of the reCaptcha road captchas I receive also have answer validation and won’t let me pass if I answer incorrectly because of a misclick.