Some outstanding questions regarding Twitter’s new country-specific censorship system:

1. Mobile clients: Will Twitter’s mobile clients also get the ability to let the user manually choose their jurisdiction, just like they now can on the full browser client? Currently neither the mobile web client nor the phone apps let you. This matters because much (most?) tweeting is done from mobile devices, especially when people are busy bringing down dictators.
2. Transparency: Twitter says it will promptly notify users if their content has been withheld, “unless we are legally prohibited from doing so.” It also says it will post requests to withhold content to the Chilling Effects website. Does that include all requests, or only those it is not prohibited from posting? Does Twitter anticipate operating in countries where it is illegal to make public the specifics of a takedown request in any jurisdiction?
3. Country-withheld content: Here is my best guess at how the country-specific censorship system works, based on testing: Before the browser requests new tweets from Twitter’s server, it first checks a cookie to see if the country location setting has been manually overridden. If it has not, then Twitter geolocates the IP address of the browser and filters the resulting twitter feed for that jurisdiction before sending the tweets along to the client. If the country setting has been manually overridden, then the browser sends along the chosen country to the server, which proceeds to filter the feed for that country, rather than the geolocation IP address. Is this correct?
4. Forms of censorship: Twitter states that sovereign jurisdictions can request the withholding of individual tweets and/or of entire accounts, and also writes that this withholding can only be reactive, in response to “a valid and applicable legal request.” The problem is, I can think of several scenarios where this might not be enough to avoid breaking local laws. For example:

• France and Germany among others prohibit search engines within their jurisdiction from linking to specific sites they deem illegal. If Twitter is not going to use a block list to pre-emptively withhold tweets containing such links in these countries, will it be breaking the law?
• If a tweet has been retweeted (natively, or by using RT, or by using quotes, or after a slight edit) by a number of users by the time a withholding request arrives which Twitter agrees to comply with, will there be an effort to remove these retweets in that jurisdiction? Might not the legal entity making the request reasonably expect these to all be one-and-the-same tweet?

It’s clear by now that Twitter’s new censorship regime is a pre-emptive move to keep the scope of censorship to within the jurisdictions of the legal authorities making the requests. This way, if Twitter is obliged by French law to remove a tweet deemed illegal in France, it will only be removed from French timelines — the rest of the world will continue to see it.

In its implementation, Twitter’s censorship system is very easy to circumvent by users — no doubt intentionally. I’ve played around with it using a virtual private network (VPN) to access my account from various countries in various setups; the workaround is trivial, albeit with a few twists.

Twitter’s own FAQ pages give two massive hints as to how to go about it:

1. Which censorship regime your account will follow is decided by having Twitter geolocate your browser’s IP address to set your “initial country” in the settings.
2. However, because Twitter might “misidentify” your country, Twitter says you get to manually override the chosen country in your Twitter account settings. Your choice is saved as a cookie in your browser. Twitter says it does not store this information on its servers.

Circumventing Twitter’s censorship is not as easy as choosing the “Worldwide” option from the dropdown menu of countries — a choice which you might assume places you outside any jurisdiction.

Choosing “Worldwide” has the opposite effect, in fact: Your censorship regime will automatically default to whatever jurisdiction your browser finds itself in. So if you choose “Worldwide” from a Swedish IP address, your country setting will immediately switch to Sweden. If you later move to an Egyptian IP address, the country setting will automatically switch to Egypt.

The “Worldwide” setting is Twitter’s default. If you haven’t changed your country manually in your account settings, this is how Twitter will choose your censorship regime. (If your IP address is not on the dropdown list of 59 countries, such as for Belgium, then the country is listed as “Worldwide” *.)

What if you are in the US, and want to ensure that your censorship regime stays American when you travel? Even though your country is listed as “United States” by default, that will change when you leave the US, unless you do this: Select any other country, save changes (and provide your password), then select the United States, and save changes. Even though the before and after settings will look exactly the same, you have now forced the browser to choose the US as your country, as opposed to whatever country you happen to be in.

But regardless of what country you happen to be in, why not choose one with best practices in free speech? A quick look at the Press Freedom Index shows that the Netherlands, Sweden and Switzerland are great choices. The governments of the Netherlands and Sweden in particular have been vocal in their defense of the net freedom agenda.

I’ve confirmed by using the same Twitter account on several browsers simultaneously across different IP addresses that the country setting for each browser is independent, saved locally in a cookie. This means you can have a second browser set to a different country, in case your default setting coughs up a censored tweet. It’s also a great way to compare and contrast censorship regimes.

In sum, circumventing Twitter’s censorship model is trivial, and I’m sure that’s not because Twitter is incompetent. What I do worry about is that this model is not robust against the future demands of censors. The wording of SOPA and PIPA, had they become law, could have been construed to classify the opt-out nature of Twitter’s censorship model as an enabler of piracy. And what about those Taliban tweets? If the US ever gets around to censoring those, it would certainly not be content with barring them just from the US; Twitter is a US company, and it can be compelled to act globally by US law.

While the newly introduced censorship model will allow Twitter to expand to countries like France and Germany, where historical baggage from World War II results in peculiar censorship regimes, or the UK, which has unique defamation laws, it is possible that new laws or future legal tests of Twitter’s approach will prohibit this censorship model. It’s great of Twitter to try, of course, but it makes Twitter’s expansion into new jurisdictions somewhat precarious, as the company may suddenly find itself faced with the grim choice of having to dismantle its opt-out censorship model in some jurisdictions, or pulling out operations from that country entirely.

Fortunately. Twitter is unlikely to ever set up shop in countries where revolutionaries are still fighting the good fight, and relying on Twitter to do so. In those countries, Twitter will not care about what the regime demands. They’ll just have to block Twitter wholesale, just as Iran, Vietnam and China currently do.

(* I suspect, but cannot prove absent a censored recent tweet to test with, that when Worldwide is selected for an IP address not from a listed country, the censorship regime defaults to that of the US. One reason I think this is the case is that only from a US IP address is it possible to select the “Worldwide” option from the dropdown menu and not have it switch automatically to the current jurisdiction, in this case “United States”. I think this is because the browser compares the two jurisdictions and sees they are the same, so does not bother to force the switch.)

Twitter’s decision to enable country-level blocking of tweets is a rational response to an Internet that long ago ceased to be that utopian place beyond location. Companies who want to grow global amid the forked legal code of today’s Internet need to follow in Twitter’s footsteps.

It would be great if there were companies that did not want to grow global, who could offer a fortified service from a free speech haven and pay no attention to the thin-skinned legal codes of the world. Such an mission would be difficult to sustain, however: Twitter is not some abstract concept; it costs money to run. Free services need advertising-generated revenue; ads require local sales teams and/or local payment systems. This means local offices, and these are within reach of local laws. A Twitter service used by the world but not paid for by the world is unsustainable. (A hypothetical premium Twitter would have the same achilles heel: local payment systems.)

In terms of fine-tuning its censorship, Twitter is catching up to Google. Google has long had the ability to censor search results on a per-country basis. It also serves mutually exclusive map datasets to India and China, where it is illegal to publish country borders at odds with the official stance. Google does this because it is heavily invested in both countries — not just with sales teams, but with development teams too. Executives face real-world criminal charges for non-compliance, as Google found out in Italy.

A highly relevant question now is: Where should the limits of tolerance lie for Facebook, Twitter and Google when it comes to censorship? When does a country, in Twitter’s own words, “differ so much from our ideas that we will not be able to exist there.”? Let’s say we’re even understanding about Germany censoring Google search for Nazi propaganda and France for Nazi memorabilia. What about the case last week of the Indonesian who posted to Facebook that he did not believe in God, and was arrested for it? Should Facebook remove the post globally? Should it remove it only in Indonesia (something it cannot currently do)? I’d say no in both cases, but many Indonesians apparently prefer not to be confronted with expressions of non-belief in their midst. (The BBC reports the page in question has been taken down, but the group still seems to be up when visited from Sweden. Facebook is able to make entire pages unavailable to specific countries.)

Is this really a fight Facebook should fight alone? Should it be YouTube’s fight to serve videos deemed insulting to Ataturk in Turkey? If we demand Facebook and Twitter and Google exit these markets rather than collaborate with laws odious to our free-speech sensibilities, shouldn’t we demand that other businesses boycott the country in solidarity? And is that really feasible?

When to tolerate censorship, then? It depends. It depends on whether a country is on a trajectory towards more free speech. It depends on whether the local laws in question are created through a broad participatory process that gives them legitimacy. It depends on whether the content objected to is an expression of a fundamental human right, such as a sincerely held belief. More cynically, it depends on whether the company in question has business interests there, chasing a growing user base. (For Google in China, this complex calculation turned against collaboration when it became obvious speech was becoming less free, not more, despite its presence.)

Today’s announcement is a bid by Twitter to ensure that excessive censorship in one jurisdiction does not bleed over into other jurisdictions. In tandem with Google, this approach amounts to a new balance of power between national jurisdictions and the web’s native interest communities. We cannot assume it is a stable equilibrium, however. One risk is that the offer of country-level content blockage is not enough for a censorious regime. It may demand that content be removed globally, else face local legal jeopardy. This is not far-fetched — demands for the removal of military “secrets” from Google Earth make no sense if they can still be seen by everyone except those within a jurisdiction. (So far, with one exception, Google has resisted such requests. China certainly tried.) It is also illegal for Google to post any information about China’s censorship requests globally, as explained in its transparency report:

Another risk is that the outsourcing of censored tweets to chillingeffects.org is only a temporary solution in a long jurisdictional arms race. Censored tweets are currently listed on chillingeffects.org/twitter, where the offending tweet can be read in full, with link and all, thus:

Since Twitter promises to alert us anytime a tweet is blocked, the Streisand effect will likely ensure wide exposure for all content that ends up there. But SOPA and PIPA were phrased to criminalize precisely this kind of “enabling” of piracy by linking, with an added extra-jurisdictional twist: Companies with a US presence would not only be enjoined from directly linking to illegal content, they would also be enjoined from doing business with non-US companies linking to it. I fear non-US legal codes will innovate to mirror this extra-judicial demand, not just for copyrighted material but for all content deemed not in the national interest. China already does (see Google’s transparency report, above).

Finally, one interesting issue I’ve not seen explained by Twitter: The mechanics of this censorship. It sounds as if there will be a block list of links for each country. I assume that once a country demands a link be put on the list, all tweets containing that link will not be shown to Twitter users in that country. But what about tweets that do not contain a link but which merely contain speech objectionable to censors? Will there be a continuously updated list of blocked terms, as is done with Chinese microblogging tools? Or will each individual offending tweet need to be flagged by censors? If it’s the latter, then there is little worry, as the half-life of a tweet is far shorter than a censor’s reaction time. But that is why China sets its own far more onerous rules for those who want to play there.

What if a country with a conservative culture or oppressive regime does demand a list of blocked terms, ostensibly to prevent obscene or defamatory speech? I suspect (and hope) Twitter decides such countries “differ so much from our ideas that we will not be able to exist there.” Twitter could then serve a full uncensored feed to users in, say, Saudi Arabia or UAE or Pakistan; the onus would be on these countries to decide if they want to invest in unilateral blocking technology of the kind China uses for its Great Firewall. That is indeed the route Iran and Vietnam have taken, and which others may yet take as Internet censorship technology gets cheaper and easier to deploy. When that kind of Internet has broadly arrived, we’ll be in the next phase in the Internet censorship arms race.