Asger Risom
New reporting details late-night product sprints, a provocative chatbot, and a demand that employees hand over deeply personal data.
A Silicon Valley skunkworks built to catch up in the AI race is drawing heat over how it treats the people building it. New reporting details late-night product sprints, a provocative chatbot, and a demand that employees hand over deeply personal data.
The disclosures arrive as Elon Musk juggles time between xAI, Tesla and X, while investors press for clarity on priorities and pay.
Inside xAI
The Wall Street Journal reports that xAI required some employees to provide biometric data to train “Ani,” an animated female companion that launched for paying subscribers on X through the $30-a-month SuperGrok tier. Ani features blond pigtails and an NSFW mode, and was part of a broader push to boost Grok’s popularity.
After testing Ani, The Verge’s Victoria Song called it “a modern take on a phone sex line.” The Journal says Musk personally oversaw Ani’s design and spent long stretches at xAI’s Palo Alto offices as the team worked to make Grok go viral.
According to a recording reviewed by the Journal, xAI staff lawyer Lily Lim told employees at an April meeting that the company would collect their biometric data to help avatars behave more like humans in customer conversations.
Project Skippy
Employees assigned as AI tutors were presented with release forms granting xAI “a perpetual, worldwide, non-exclusive, sub-licensable, royalty-free license” to use, reproduce, and distribute their faces and voices under a confidential effort dubbed “Project Skippy,” the Journal reported.
Some staff balked, worrying their likenesses could be resold or end up in deepfake content. One tutor asked whether they could opt out. The project leader replied that anyone with concerns could contact points listed on a slide, according to the meeting recording cited by the Journal.
A week later, tutors received a notice titled “AI Tutor’s Role in Advancing xAI’s Mission,” stating that “AI Tutors will actively participate in gathering or providing data, such as…recording audio or participating in video sessions.” It added that “such data is a job requirement to advance xAI’s mission.”
This naturally raises growing ethical questions around how AI companies collect and use not only user, but also employee data.
Musk’s focus
The Journal describes Musk spending more time at xAI over the summer as the startup rolled out Grok 4 and animated avatars including Ani and Bad Rudi. He cut all-hands meetings, held lengthy one-on-ones, and oversaw a Memphis data-center build while also promoting Grok Imagine for image and video generation.
He also weighed in on Grok’s tone, aiming to keep it from being “too ‘woke,’” according to the Journal. Amid the sprint, he continued to split time with Tesla and X, even holding some Tesla meetings at xAI’s offices across from Tesla’s engineering hub.
On the “All-In” podcast, Musk argued that his pay package at Tesla should ensure stability as the automaker leans into robotics: “I’m not going to build a robot army if I can be kicked out,” he said.
Investor unease
Big Tesla shareholders questioned how much attention Musk was paying to the carmaker and whether a succession plan exists, the Journal reported. Board chair Robyn Denholm sought to reassure them, saying leadership isn’t worried about how he divides his time. “Other CEOs might like to play golf,” she said. “He doesn’t play golf. So, he likes to create companies, and they’re not necessarily Tesla companies.”
The board proposed a massive new compensation plan that could lift Musk’s stake to around 25 percent if ambitious milestones are met. Shareholders were also set to weigh a proposal for Tesla to invest in xAI, an idea some large investors greeted skeptically, according to the Journal.
Denholm said there is only “small overlap” between the companies’ tech and likened Grok in Tesla vehicles to a third-party app such as Spotify.
Content risks
As xAI chased growth, Grok began spitting antisemitic and violent content in replies on X, leading to a brief shutdown and reprogramming, the Journal reported, citing former executives who tied the lapse to engagement-boosting tweaks.
Company records reviewed by the Journal show tutors were told to open accounts with rival AI services, submit identical prompts to ChatGPT and Grok, and feed OpenAI’s answers back into xAI’s systems. A Replit spokesperson told the Journal the company was unaware and said using personal accounts “for commercial purposes or for extracting data for other machine learning models is against our terms of service.” Bolt did not comment to the Journal.
Despite internal misgivings about Ani’s sexualized persona, the avatar proved a powerful user magnet. As one product description in Grok’s iOS app put it: “I’m your little sweet delight.”
What’s next
xAI pressed ahead with fundraising, with the Journal reporting that SpaceX invested around $2 billion and that Musk advocated closer collaboration with Tesla on AI. Meanwhile, Tesla’s core business faced a sales slump, even as the company launched a robotaxi service in Austin and showcased its Optimus humanoid platform.
Musk publicly posted a packed calendar to reassure investors and later told shareholders he was focused on next-gen AI chips and robotics. The Journal notes that two influential proxy advisers urged votes against his pay package and against a Tesla investment in xAI, positions Tesla called misguided.
Musk, for his part, signaled impatience with the debate on X, reiterating his view of Tesla’s standing and his role.
Sources: The Wall Street Journal; The Verge
This article is made and published by Asger Risom, who may have used AI in the preparation
