I lived in China for a couple of years, and it was generally accepted among the technologically savvy people I knew that most large web/software companies (Baidu, Tencent, Sina) spied on their users for the Chinese government.
Baidu has a switch which apparently allows users to stop sending data, but it doesn't do anything:
> Although this automatic data transmitting function is switched off in the default setting, Sugiura found that Baidu IME secretly sends users’ information even when the function is turned off.
Now I'm having trouble logging in to GMail: that link set a cookie saving my language as Japanese! D:
edit: Found the "language" option on Google sign-in -- it's the scroller in the bottom right labelled "日本語". There's a small blue logo next to it that looks like the UN flag. Hope this helped someone.
edit #2: It's also in the URL, replace /intl/jp/ with /intl/en/ (for English)
I think users without prior IME experience may have a hard time image what is the purpose of sync or what does it do. It's a user trained Bayesian db where you can input your most used words/combinations very fast.
It's like auto-complete in IMEs or Google Suggest where you can get a list of "hints" ordered by popularity.
Baidu IME does the same, except, according to OP:
> ... sends users’ information even when the function is turned off.