For UTF-8 adding ensure_ascii = False to json.dumps would fix it. For HTML, there is no simple way as far as I know. With some searching you can find some workarounds. Or you can consider using https://github.com/simplejson/simplejson
I did point out this issue almost a week ago https://gerrit.wikimedia.org/r/#/c/119637/4/i18n/qqq.json -Niklas