Is China censoring phone conversations?

By Danny O'Brien/CPJ Advocacy Coordinator on March 24, 2011 11:02 AM ET

Are Chinese mainland citizens, as has been reported, finding their telephone conversations cut off whenever they mention the word "protest?" While large-scale, real-time voice recognition is a technological possibility, it is at the edge of what is believed likely. It would certainly be revealing about the capabilities of the Chinese government if these anecdotes proved to be widespread. 

Such a powerful capability is generally seen as something Chinese authorities would want to keep confidential, rather than reveal in simplistic blocking strategies such as single-word phone hang-ups. That said, it is entirely understandable that technology users in China would believe such behavior from their telephone system providers. Much of China's Internet blocking works in exactly this way: You use the wrong term on a search engine, and your connection to that search engine mysteriously shuts down for a few minutes.

In the last month, users of Google's Gmail service have reported random interruptions and slowdowns in service. Many of them complained to Google, assuming that the breakdown was on Google's end. In fact, Google and others report that the slowdown came from within China, and was deliberately targeted at the service. Christine Chen, a Google executive, told the IT news outlet IDG News Service: "There is nothing technically wrong on our side, so you will have to ask the [Chinese] government as it is clearly an issue on their end".

Those using commercial virtual private networks (VPNs) to bypass such censorship have also found their connections mysteriously fail. As Oiwan Lam at Global Voices Advocacy reports, the Great Firewall has, in some places, been expanded to block the Internet protocols that most VPNs use. Blocking at such a low level can look like a failure of client software, rather than a deliberate act of governmental interference.

Such blocks are far easier to implement than telephone voice recognition. They're also relatively easy to circumvent. But the constant cat and mouse game between those who are blocked and those who seek to bypass those blocks serves to remind everyone that the Chinese authorities are always watching.

Some governments are happy to present their censorship openly (even jocularly, as with Qatar's blocked page). But China mostly prefers to implement its censorship in the shadows. The self-censorship of Chinese social sites is cloaked in euphemistic language, as this slide-show of recent censorship messages collected by CPJ board member and New America Foundation senior fellow Rebecca MacKinnon shows. When the Tibetan affairs site TibetCul was closed earlier this week, the server owners refused to tell the operators of the site why it had been disabled, except that "higher authorities" had ordered the closure.

High-tech censorship is rarely perfect, but it does not have to be. The arbitrary and random nature of China's information blockade has the advantage for the Chinese authorities of making everyone paranoid of its true extent or who will be next. Does China have the technology to recognize and block the word "protest" however you say it? Probably not. But if believing that means that you are less likely to write about any protests, or talk to others about them, it is effective censorship nonetheless.


Google Voice has done real-time transcription of phone calls and voicemail messages for a couple of years already, so why couldn't China do similar on all incoming/outgoing international calls? Surely the computer hardware isn't too expensive for the world's second-largest economy.

Hey Kevin,

It's absolutely true the Golden Shield is extremely well funded, and as I said, it is technologically possible. But it *would* cost a lot, and require a lot of CPU that might be otherwise put to other purposes. There's always an opportunity cost even for well-funded projects.

Also, there are lots of things that are technologically possible that we don't have evidence for. Right now, the assertion is that we do have evidence for this particular behavior, and the evidence are two anecdotes about phone calls dropping calls (actually domestic internal calls not international ones) after somebody said the word "protest".

This is actually testable: we can get people to say the same words and see if their call is dropped, and also ask more widely to see if anyone else has seen the same chain of events. I've yet to see anyone say "this happened to me", and plenty of people saying "I tried this and nothing happened".

It's important when trying to quantify the truth of anecdotal data to also take into account how compelling the anecdote itself is. If it's a great story, the tale will propagate and stick in people's minds even if it isn't universally true or widespread. The "protest" keyword story is a great and visceral depiction of the level of censorship currently experienced by Chinese mainland tech users, but the truth is that the probability of a call dropping is so high that two connections like this don't make for a causal connection with surveillance, no matter how compelling it is.

In almost every context, this doesn't really matter. It was, after all, just the lead-in to a great article that spelled out a lot of testably true indications of the current high level of Chinese censorship.

But part of my job is trying to accurately judge the current practices and capabilities of states who practice surveillance and censorship on journalists. Turning on such a feature, and then demonstrating that capability so publicly would be an extreme ramping up of China's abilities as well as a very revealing indication of what they can do, so I'm naturally very interested in determining whether it's a valid indicator. Right now, everything points to this being a great story with little evidential value.

Note that I could be wrong! And I'm going to keep looking at stories like this to see if we can get a better indication of whether this is widespread or not. What we *do* know, however, should be disturbing enough. Blocking keywords, taking down websites, sending journalists targetted malware (as just happened to Al Jazeera's Melissa Chan ). These things require far fewer resources, but are equally threats to press freedom.

Here's an update by the New York Times, taken from

An article on Tuesday about Chinese censorship of digital communications began with a description of two interrupted cellphone calls, which were cited as possible examples of “a host of evidence over the past several weeks” that the authorities were increasing their efforts out of concern that antigovernment sentiment might spread from Arab countries. In one call, a Beijing entrepreneur lost his cellphone connection after he used the English word “protest” twice. In the second, a call was lost after the speaker twice used the Chinese term for protest.

The article did not point out that in both cases, the recipients of the calls were in the Beijing bureau of The New York Times. Because scrutiny of press communications could easily be higher than for those of the public at large, the calls could not be assumed to represent a broader trend; therefore, those examples should not have been given such prominence in the article.

That makes a lot more sense (although it's still anecdotal). Here it is clearer that the perceived blocking was targeted at a particular high-risk group: journalists, as a matter of fact. In these cases, it could just as well have been down to human surveillance, rather than any high-tech censorship. Widespread voice recognition costs billions, but tapping a bunch of foreign correspondents is neither costly nor unknown.

In the New York Time's defense, their original article never explicitly stated that this monitoring was automated nor widespread. But that was certainly how most people seem to have interpreted the story.

NEWSFLASH: China interference prompts postponement of controversial IPO
Chinese pressure upon global financial institutions has forced the postponement of the initial public offering (IPO) of China internet, content and technology company CRS, according to Director and Founder Mao Ze Wrong.
CRS, operator of the satire website,, and the Chinese-centric search engine,, had planned its IPO for this week, amidst increasing interest in its reach across China and unusual content.
Mr Mao made the announcement Friday in Hong Kong, where he had spent the past two weeks meeting with bankers and financial investment houses. “This is not a cancellation, only a postponement,” he said. “We will still go ahead with the IPO, but in a few months.”
China watchers say there has never been a site so critical of China to seek an infusion of investment funds, but acknowledge that the rapid growth of China has also prompted a massive expansion in research, retail and media, including many critical of China’s record on human rights, freedom, censorship and spitting.
For more details, see:

Social Media

View All ›