New York legislators proposed requiring AI companies to warn if their chatbots might cause harm, but facing heavy lobbying, they backtracked.