For some time now, Twitter has been trying to curb abuse and harassment on the service, after admitting that it hasn’t done enough to stop trolls and deliberate misbehavior in the past. The latest step the company is taking is to throttle the reach of users it believes are being offensive or targeting others for abuse.
Over the past few days, a number of Twitter users have reported getting warning messages from the company saying: “We’ve temporarily limited some of your account features.” The warnings come with a time period during which the restriction will be applied, a time that appears to be set at 12 hours in most cases.
“Creating a safer environment for people to freely express themselves is critical to the Twitter community,” the warning reads. “So if behavior that may violate the Twitter Rules is detected, certain account features become limited.”
The restriction effectively limits the reach of individual accounts to users who already follow that person — their tweets are essentially hidden from everyone else for the period of time shown in the warning. After the time-out period expires, all of the normal features of their account are restored, the message says.
Get Data Sheet, Fortune’s technology newsletter.
According to a report by BuzzFeed, if one of that user’s followers retweets something they posted, that retweet will also be subject to the time-out limit, and won’t be seen by anyone other than users who already follow that account.
Although some users who have been hit with the restriction have said they believe it was because they used an abusive term, a spokesperson for Twitter told BuzzFeed that when deciding who to apply the limit to, the service’s harassment and abuse team looks at a pattern of overall behavior, rather than just individual words.
Twitter’s crackdown on abuse started after CEO Jack Dorsey announced that the company was committing itself to tackling the problem, and also to doing so transparently and through an “open and real-time dialogue” with users.
We're taking a completely new approach to abuse on Twitter. Including having a more open & real-time dialogue about it every step of the way https://t.co/a1SV7URPEK
— jack (@jack) January 31, 2017
The company’s new-found religion on harassment may have also been sparked by the fact that at least two of the companies that were rumored to be considering an acquisition of Twitter last fall—Salesforce and Disney—reportedly shelved the idea in part because of the environment of abuse that exists on the platform.
Almost since Twitter started in 2007, the company has struggled with the two sides of its open approach to free speech. The company has boasted about being the “free-speech wing of the free-speech party” and talked about how the service helped free Egypt during the Arab Spring. But at the same time, its hands-off approach arguably encouraged abuse.
Some right-wing websites have already started criticizing the company’s latest move as a crackdown on free speech, and an attempt to enforce “politically correct” behavior.
As The Verge noted in its report on the new anti-abuse measure, the kind of restriction that Twitter has implemented with its latest move has a long history on the social web, where it is often referred to as a “shadow ban.”
In a number of online forums and other web communities, abusive users were limited in this way, but in most cases they were not told of the restriction—so they would post and behave as normal, but no one would see their messages. Twitter has taken the added step of informing users that they have been restricted.
Letting users know about the time-out risks triggering a backlash. But even when stealth bans have been employed, users almost always found out about them anyway, either from friends or by noticing that they weren’t getting as many replies as usual. So a backlash was inevitable. And by limiting the amount of time spent being throttled, Twitter shows users that the restriction isn’t permanent.
Among the other steps that Twitter has taken to curb abuse is a new filter that tries to weed out harassing or “low quality” replies to user’s tweets. These replies are grouped together underneath a tweet and a user has to click in order to see them.
But Twitter has made some recent missteps on the harassment issue. The company announced on Tuesday that it would no longer notify users when they were added to a list (a way users can keep track of multiple accounts in one place), but this sparked an outcry from a number of users who said adding someone to a list is often a prelude to harassment, and that they would like to know about it.
Within hours, Twitter announced that it had made a mistake, and that the notification change would be reversed. To some users, the speed with which the company acknowledged the error and fixed it suggested that Twitter does care about the abuse and harassment issue, and is watching closely to see how its new measures are being received.
