The popular Twitter bot @fuckeveryword was working its way through an English dictionary from “fuck a” to “fuck zyzzyvas.” But it hit a fraught stretch of the “N” section right as Twitter kicked off a purge of extremist and white supremacist accounts, posting a particularly vile string of racial insults including the N-word. Soon afterward, Twitter suspended the account.

It should have been easy to spot trouble ahead. The creators of the bot uploaded a list of every word based on a wordlist created by Christian organization SIL International, which included the one that seemingly got it kicked off Twitter this week. The account had previously attracted criticism for programmatically insulting, among other things, “Allah” and “jews.”

One of the creators of @fuckeveryword, who identified himself as George Miller during an interview over Twitter direct message, told me that the account had already received several warnings in the year prior to being shut down, and had been temporarily suspended after one particularly hateful tweet.

“I have yet to see a school ban the dictionary, but, given the political climate, I suppose anything is possible,” said Miller, who said he lives in Philadelphia and has a nascent career as a SoundCloud rapper.

It’s not clear whether Twitter has ever suspended a bot for hate speech before, but it could be a sign of things to come. State-sponsored botnets on Twitter are widely suspected to have used the platform to sow chaos, and until recently the site’s executives have opted for a hands-off approach about racial extremism on the site.

Fuck Every Word’s ban appeal has been denied. The bot was promoting violence, apparently. pic.twitter.com/dCEBCBGV8U

— hawk (@hawktherapper) December 22, 2017

But Miller defended the @fuckeveryword bot by arguing that it would undercut the concept to scrub certain words from its vocabulary.

“In this case, while I’m sure the presence of such words is deeply offensive to some people, the context is very clear,” he said. “Racial slurs are a part of the language. There were complaints for many other racial epithets, religions, sexuality slurs and my stance is the same.”

In practice, that defense tends to make online spaces uncomfortable for marginalized groups. Miller also conceded that he’d gone as far as adding words to the dictionary — resulting in zingers, he said, like “fuck anime” — for comic effect, so it’s not entirely clear why he refuses to remove particularly egregious language.

Still, he bristled at the suggestion that @fuckeveryword had given visibility to racist rhetoric. “Hopefully we aren’t associated with Nazis,” he said. Earlier in the “N” section, he pointed out, the account tweeted: “fuck nazism.”

@Fuckeveryword is just a bot, but bots left to their own devices often veer into hate speech. Other bot makers take a different approach. Serial Twitter bot maker Darius Kazemi made an open source word list of slurs his bots can’t say. He told Motherboard that this tool, called wordfilter, “makes sure that none of my bots is ever going to say the n-word. It's just a baseline thing.” “It's easy to make it so that robots never say slurs,” another bot maker, Parker Higgins, told Motherboard. “You just add one more line.”