Elon Musk wants to make Grok AI an option for X premium users to compose tweets

If you still use X, you might have noticed some AI features seeping into the platform lately, powered by Elon Musk’s xAI chatbot, Grok.

Last week, X rolled out “real-time customized news for you created by Grok AI.” In the pre-Musk Twitter era, there used to be something called a “curation team” that identified trending topics and curated the platform’s Explore page. Trending topics were whatever users were talking most about: the latest iPhone, Pi Day, or more pressing subjects like mass shootings and protests. Musk laid the entire curation team off shortly after acquiring the company, and now, many months later, it’s been replaced with AI. That’s going as well as one might imagine.

After an earthquake shook New York City last week, X’s Grok-powered trending tab dredged up jokes from users about the city’s mayor, Eric Adams, sending 1,000 police officers to the Earth’s core—and pushed it as real news. Even worse, the AI trending tab ran a headline last week that read, “Iran Strikes Tel Aviv with Heavy Missiles,” which was completely made up.

Despite the pitfalls, X’s Grok-ification shows no signs of slowing. One source at X told me that Musk has directed engineers to add Grok to the tweet box for users who pay for X Premium, the platform’s subscription feature, which means that users may soon be able to tweet with the help of AI (I imagine this will look similar to what X CEO Linda Yaccarino already posts). The source said Musk “wants people to sound smarter,” and he thinks Grok could help with that. For instance, after being prompted about outer space or popular memes, AI could assist with creating posts about those topics.

But making those posts “smarter” is a tall ask. Grok, after all, is trained on X’s huge years-old archive of user posts, and many of those aren’t exactly erudite.

What Musk wants AI to do with posts on X isn’t exactly novel, either. While typing an email over Outlook or Gmail, users are often prompted with options to autocomplete their sentences. LinkedIn also offers options to write using AI. I’d say the difference between firing off emails and posting on X is that the latter tends to involve more creativity—a crafted personality, or perhaps personal authenticity. AI, as it stands today, is devoid of all of the above.

There is also the nontrivial problem of spam, which engineers seem unsure of how to tackle, and Musk seems unconcerned about, according to the source. There is so much spam on X that, when Musk posted that xAI would open-source Grok, whoever runs Grok’s X account appeared to poke fun at the widespread spam issue by mimicking the style of popular porn bot messages that pollute the platform.

Musk asked engineers to add Grok to the post-composing tool a few months ago, and, according to the source, the team is stalling. What’s more, the source says the xAI API is slow, making it hard for the X team to work with, and naturally, no one thinks it’s a great idea to let users use AI to fire off shitposts, or aggressive nonsense, at a higher volume.

As someone who has covered X since it was still known as Twitter, I think integrating Grok into X is an easier and quicker win for Musk than, say, creating a whole payments tool, which has long been in the works. If engineers fail to dodge the AI directive, well, I’ll be interested to see how users put Grok to work as their ghostwriter.

Kylie Robison

Want to send thoughts or suggestions to Data Sheet? Drop a line here.

The rest of today's Data Sheet was written by David Meyer.

This story was originally featured on Fortune.com