The latest incident to highlight this tension is Twitter’s blocking of a number of accounts that preserve the tweets of politicians, as a way of tracking their public statements about social issues. Twitter first blocked the U.S.-based account @Politwoops in June, and now it has blocked a number of other similar accounts in different countries that were run by the Open State Foundation, a non-profit group that promotes transparency through open data. According to a blog post at the Foundation site:

“Twitter said that its decision to suspend access to Politwoops followed a ‘thoughtful internal deliberation and close consideration of a number of factors’ and that it doesn’t distinguish between users. Twitter wrote: ‘Imagine how nerve-racking – terrifying, even – tweeting would be if it was immutable and irrevocable? No one user is more deserving of that ability than another. Indeed, deleting a tweet is an expression of the user’s voice.’”

Twitter’s rationale appears to be that the rights of the individual politicians whose tweets are preserved by these accounts outweigh any social benefit that comes from such a service. But it didn’t always feel that way: When Politwoops first started in 2012, the service was notified that keeping deleted tweets was a breach of the terms of use, but after the Sunlight Foundation explained the purpose and said it would be judicious about what it chose to keep, Twitter approved it.

What the company seems to be saying is that politicians are Twitter users just like anyone else, and since regular users can delete their tweets if they wish (something that appears to be an emerging trend in some quarters), then why shouldn’t an MP or congressman have the same ability? Since it is also concerned with a slowing growth rate, Twitter probably also doesn’t want new users to get the sense that their tweets are permanent or might some day be used against them.

Twitter's position here is ridiculous. Public statements *are* immutable and irrevocable. That's the point of them. http://t.co/CxZOx7wRGz

— Ivor Tossell (@ivortossell) August 24, 2015

At the same time, however, there’s a clear social value in having tweets about important political topics preserved, in the same way that there’s a social value in recording off-the-cuff remarks made by politicians at meet-and-greet events. That kind of thing got both Barack Obama and Bill Clinton in trouble when they made comments that were reported by “citizen journalist” Mayhill Fowler, who was working for The Huffington Post.

Politicians are also — at least in the U.S. — seen as being more than just your average user of a social network or public service. They are considered to be public figures, and therefore the press in particular is given much more leeway in reporting on their personal behavior.

Twitter may be a multi-billion business that wants to boost its flagging share price, but it is also a company that has staked a large chunk of its reputation on being a platform for free speech. In addition to the comments about being the free-speech wing of the free-speech party and a digital town square, Twitter founders including Evan Williams have written in the past about how the “tweets must flow” — in other words, how the company didn’t want to bow to demands to restrict speech.

So @Twitter is out here suppressing speech. OK then.

— #TenureTrackHustle (@meredithdclark) August 24, 2015

The service has done its best to walk this line in the past, by implementing a fine-tuned form of censorship based on geography: Using its so-called “Country Withheld” tool, Twitter can make sure that tweets about Nazism aren’t seen by users in Germany (since these kinds of statements are illegal), or that specific tweets that are the subject of a court action in Turkey aren’t visible to people in that country. And it has fought court cases trying to uphold the right of users to post information about things like WikiLeaks.

In this case, however, Twitter has decided that the social value of a politician’s tweet isn’t enough to maintain access to those tweets over the user’s wishes. The move may also have been sparked by the opening up of Twitter’s search API, which now provides access to every tweet that has ever appeared on the service since it started in 2006.

The problem is that it’s easy to see any number of cases where there would be an obvious and compelling public interest in maintaining access to a politician’s statement about a public issue. If a U.S. senator or British MP says something that indicates his or her true feelings about an important topic, isn’t that the same as making a comment in public at a political event? If those individuals can’t make the latter disappear, why should we allow them to remove the former? As Arjan El Fassed of Open State said:

“What elected politicians publicly say is a matter of public record. Even when tweets are deleted, it’s part of parliamentary history. What politicians say in public should be available to anyone. This is not about typos but it is a unique insight on how messages from elected politicians can change without notice.”

For some, the company’s decision on Politwoops feels similar to what Google is being forced to do by implementing the so-called “right to be forgotten.” In one particularly Orwellian case, the company is being asked to remove even the mention of the fact that something has been removed. If we are to allow anyone the opportunity to “un-say” something, then what is the point of having public information networks and services like Twitter and Google in the first place?

As a number of people have pointed out, it will still theoretically be possible to keep track of specific politicians’ deleted tweets by scraping their Twitter feeds. All Twitter has done is make it impossible for specific accounts to do so using its API (application programming interface). But it’s still worth questioning the default decisions made by platforms like Twitter and Facebook, since so much of our public speech takes place on them. As the Sunlight Foundation put it:

“Twitter’s decision to pull the plug on Politwoops is a reminder of how the Internet isn’t truly a public square. Our shared conversations are increasingly taking place in privately owned and managed walled gardens, which means that the politics that occur in such conversations are subject to private rules.”

You can follow Mathew Ingram on Twitter at @mathewi, and read all of his posts here or via his RSS feed. And please subscribe to Data Sheet, Fortune’s daily newsletter on the business of technology.