Enlarge /. Sundar Pichai, CEO of Google, speaks in 2017 in Wuzhen, China.
YouTube says it has introduced "a solution" to a "bug in our enforcement systems" that automatically deleted comments containing two sentences that the Chinese government criticized. However, during an email exchange and phone call with Ars Technica, a company spokeswoman declined to provide real details about why the YouTube software deleted the comments in the first place.
As I explained on Tuesday, "共匪" means "communist bandit". It was a derogatory term used by nationalists during the Chinese Civil War that ended in 1949. It continues to be used by Chinese-speaking critics of the Beijing regime, including in Taiwan.
"五毛" means "50 cent party". It is a derogatory term for people paid by the Chinese government to take part in online discussions and promote official positions of the Communist Party. In the early years of the censored Internet in China, such commentators were reportedly paid 50 cents (in China's currency, the yuan) per post.
By Tuesday, YouTube automatically deleted all comments that contained these sentences. I confirmed the behavior myself on Tuesday morning. Comments that include either phrase will disappear in less than a minute, while other comments – including those that contain other Chinese phrases – will remain on the website.
Users have been reporting this behavior since the end of last year without YouTube responding. That changed on Tuesday when high-profile news sites – starting with The Verge – reported on the story. YouTube had fixed the bug within 24 hours of the release of The Verge story.
And YouTube says it was a mistake, not an intentional political decision. But not everyone is convinced.
"This alleged" mistake "follows a long, troubling pattern of content censorship by Google to gain favor with the Chinese Communist Party," Sen Josh Hawley (R-Mo.) Wrote in a Wednesday letter to the Google CEO, Sundar Pichai. Hawley is one of many who suspect that this was an intentional political decision – not just an innocent mistake.
The argument for transparency
On Wednesday I exchanged emails and called a YouTube spokeswoman. She seemed ready to help, but couldn't offer me many details. She said that YouTube relies on classifiers to decide which comments to delete, and that YouTube classifiers don't take "the right context" into account. She said she couldn't give more details.
I'm sure it wasn't her fault. In a large company like Google, decisions about what to tell the press are made by the people who actually speak to reporters like me. But I think Google as a company is making a mistake by keeping it so secret.
It seems plausible that there is an innocent explanation for YouTube's mistake. For example, the terms "五毛" and "共匪" often appear in heated arguments that include other abusive (but less political) Chinese terms. It is easy to imagine that an algorithm would classify it as abusive without considering the political consequences – and without a human employee at Google realizing it.
Alternatively, people on the Chinese government payroll may have figured out how to play YouTube comment moderation rules by tagging millions of comments that criticize the Chinese government. Or maybe a low employee with sympathy from the Chinese government has included the sentences in a list of prohibited sentences without the knowledge or consent of their superiors. In a company where management mostly doesn't read Chinese, it's easy to miss.
Each of these statements seems both more plausible and less harmful than senior Google managers who deliberately choose to censor phrases to promote Beijing's favor. However, if Google refuses to be transparent about how and why this error occurred, many people will assume the worst.