Elon Musk’s xAI Is Becoming a Leaky Ship

In recent weeks, a series of leaks has plagued Elon Musk’s xAI. Indeed, a slow, steady drip of stories from major news outlets has found, as its basis, a healthy helping of anonymous sources and internal company messages that said sources seem more than willing to share. The portrait of the company that’s being painted is one of a tumultuous startup, where change is rapid and much of the staff has been moved around or let go.

The most recent leaks were reported in a New York Times piece on Friday, which, among other things, details a company meeting held on Wednesday involving Musk. At the meeting, Musk reportedly engaged in his usual habit of grandiose verbiage, which always seems designed to position the billionaire and his companies at the central pivot point of the universe:

“We are the only company where the mission is truth,” Mr. Musk told his workers as part of an hour-and-a-half presentation that was listened to by The New York Times. “If you force the A.I. to lie or believe things that are not true, you’re at great risk of creating a dystopian future.”

However, the narrative that the Times piece advances is of a company that is much more interested in trying to gain social media attention and up its user base than pursue a “mission” that is “maximally truth-seeking”—whatever that’s supposed to mean. By quoting a large number of anonymous sources, the newspaper purports to give context to several controversial xAI incidents that have occurred over the past several months.

For instance, remember when Musk insisted that Grok, his supposedly anti-woke chatbot, was too woke? The Times confirms what many already suspected:

In May, Mr. Musk told more than 100 employees in a group chat that Grok was too woke, according to three people familiar with the discussion.

Later that night, an engineer changed Grok’s code in response. The change caused Grok to begin bringing up South African politics when answering completely unrelated questions on X. It falsely insisted, for example, that the country was engaging in “genocide” against white citizens.

The Times provides fresh details on what Musk, himself, has already openly admitted multiple times: his engineers have been monkeying with Grok to “fix” it every time it gives an answer Elon doesn’t like. Another incident, in which Grok went MechaHitler, was reportedly spurred by Musk, who is alleged to have “seized control” of the chatbot from several researchers and given the reins to two other staffers, who then “attempted to make Grok’s responses edgier, aiming to draw more attention so the chatbot would go viral on X.” The bot spewed antisemitism until devs intervened.

The MechaHitler incident notably got Grok back in the news and made people pay attention to Musk’s company (although it goes without saying that, in many cases, the attention was negative).

The Times also reports that the recent rollout of xAI’s sexy chatbots as part of Grok Imagine “frustrated some of the company’s researchers who thought xAI was straying from its scientific mission.” Those researchers ultimately “told colleagues they were leaving xAI because of these bots,” the newspaper writes.

The Times article represents only the latest of several instances of sources close to the company gabbing to journalists. Just yesterday, the Wall Street Journal quoted a number of sources who claimed that some of the recent turnover at the company was spurred by internal disputes about management styles and company finances. Alex Spiro, Musk’s attorney, disputed much of the report. Gizmodo reached out to Spiro and xAI for comment. We’ll update this post when we receive a reply.

About a week ago, Business Insider cited internal company messages to report on layoffs at the firm. According to the outlet, around 100 staff members were let go. Around the same time, the outlet cited internal messages to report on an internal shake-up of the team that is supposed to train Grok. That shake-up reportedly led to the departure of nine “high-level employees” on the human data management team.

In short, the tumult at Musk’s AI company would appear to be spilling out into public view thanks to those current or former staffers who are not exactly thrilled about the direction of the business. The word one source used to describe a recent company meeting was “chaotic.” That sounds about right.