Bing Bests Baidu Censorship
Abstract
Independent research from Xia Chu has shown that, in addition to non-China content, Bing censors a vast amount of content that is hosted inside China and which is not censored by China-based internet companies like Baidu. After communicating our issues with Microsoft, Bing removed certain censorship rules (kudos to Bing), but much work remains to be done.
We recently called for Microsoft to release its transparency report for Bing (as have others - full disclosure, Rebecca sits on our advisory board). Microsoft has yet to respond to this request. But Xia’s independent research of Bing’s China censorship policy could be regarded as a de facto transparency report for the search engine.
In this thorough study, the results of which we have verified, Xia examined Bing's SERP (search engine results page) for over 30,000 sensitive and nonsensitive query terms, and launched these queries from both inside and outside of China. Comparing and examining these results, plus querying with special search operators, reveals unprecedented detail on Bing's China filtering practices.
The main findings from Xia’s research include:
-
Bing has a list of “forbidden” terms where no results are shown. 139 such terms have been identified.
-
Bing has a blacklist of websites that it never shows to China users. 329 such websites are identified. (5 have been lifted after our communication with Microsoft.)
-
Bing has a huge blacklist of URLs that it never shows to China users. 1593 URLs are identified in this study, which is only the tip of the iceberg.
In particular, we’d like to point out two important issues:
-
Bing's filtering is very broad. For example, Bing blocks all results from five foreign language editions of Wikipedia: Japanese, French, German, Dutch and Swedish. Many URLs on the New York Times website are censored, both for sensitive topics such as Zhou Yongkang and Tibet but also for non-sensitive pages about real estate and sports. (Note, these five Wikipedia sites are no longer censored after we approached Microsoft about this issue.)
Note: even though the censorship notice says “部分结果” (meaning “some results”), the SERP is actually empty and shows no results.
-
Bing has extensive filtering rules for China-based websites like blog.163.com, baidu.com, blog.sina.com.cn, 360doc.com, ifeng.com, douban.com and others. The content filtered by Bing still appears on these domestic websites and is not even being filtered by China-based search engines like Baidu.
Below are screenshots from Bing and Baidu for the site heweifang2009.blog.163.com, which belongs to Professor He Weifang, a prominent pro-democracy legal scholar at Peking University. His blog is not censored by the domestic blogging site blog.163.com, which is operated by NetEase, a major Chinese internet company. This website is not censored by Baidu either, but is being censored entirely by Bing in China.
Note: even though Bing’s censorship notice says “部分搜索结果未予显示” (meaning “some parts of search results are not shown”), the SERP is actually empty and shows no results.
Based on Xia’s findings, our conclusion is that Bing has not lived up to their promise of "support for freedom of access to information". For example, removing five foreign language editions of Wikipedia, does not support Bing's earlier claim: "if we are required to implement (a censorship) request, we will do so narrowly". Also, the fact that Bing filters vast amounts of content that is hosted in China, and significantly more than Baidu does, is saddening and inexcusable.
Before we published this post, GreatFire contacted Microsoft. In particular, GreatFire raised the issue that Bing censors domestic content which is not being censored by Baidu, and also raised the filtering of the five Wikipedia sites. Microsoft replied:
Generally, it is Bing’s policy is to remove results as narrowly as possible to comply with Chinese law. Bing regularly reviews removed results and our related processes to ensure that prior removals due to legal compliance are still valid. We will actively resolve any issues we identify that do not comply with our policies, including issues that match the scenarios you described. We appreciate work done by Greatfire, Xia, and others. Such feedback enables us to review and validate our own processes. We commit to addressing these issues comprehensively and expeditiously.
To our delight, Bing has lifted the ban on the five Wikipedia sites in China. This demonstrates that when filtering rules are revealed (as with a third-party Transparency Report), corporations pay attention to these issues more seriously and take corrective measures.
However, Microsoft still has not taken our message seriously enough. Out of the 329 site-level filtering rules, Microsoft only removed these five. All of the other filtering rules remain effective as of time this post was published, including broad censorship and heavy-handed censorship of domestic content.
We make the following requests to Microsoft:
-
Explain why Bing censors so much content, including domestic content that is not even censored by China-based web properties like Baidu.
-
Follow your action regarding the five Wikipedia sites by examining and removing other inappropriate filtering rules for users in China.
-
Audit the actual procedure and handling of censorship in China, to verify that all government requests are scrutinized by proper personnel and to share the results of this audit. In particular, please disclose the internal processes and approval mechanisms that are in place when Microsoft receives a request to censor domestic content. Also, please conduct and publicize a post-mortem investigation as to why the entire Japanese, French, German, Dutch and Swedish Wikipedia sites were censored in the first place.
-
Not attempt to hide censorship details, e.g. by altering censorship notices, altering how search operators like “site:” and “url:” work, or tweaking ranking algorithms.
The full Bing censorship rules identified and linked to from the research paper can be found on the master spreadsheet. Below is a summary of filtering rules Bing implements in China for domestic content. Only rules identified in this study have been counted although we believe the actual rule counts are much higher.
Domain |
# Filtering Rules |
Comment |
blog.163.com |
43 |
NetEase blog |
360doc.com |
19 |
a file storage and sharing service |
56.com |
2 |
a video hosting site |
baike.baidu.com |
31 |
Baidu baike (a Wikipedia knock-off) |
hi.baidu.com |
2 |
Baidu blog |
tieba.baidu.com |
5 |
Baidu discussion forum |
wenku.baidu.com |
5 |
Baidu library |
zhidao.baidu.com |
7 |
Baidu question/answer service |
douban.com |
29 |
Douban, a review site |
ifeng.com |
17 |
a major portal site |
qq.com |
19 |
a major portal/personal blogging site |
blog.sina.com.cn |
26 |
Sina blog, China's biggest blogging site |
baike.soso.com |
10 |
Soso Baike (a Wikipedia knock-off) |
bbs.tianya.cn |
5 |
Tianya bbs, a major forum site |
tudou.com |
4 |
Tudou, a video hosting site |
youku.com |
12 |
Youku, a video hosting site |