New creepy details are emerging about how Musk really built Grok’s anime girlfriend Ani.
On the day Elon Musk secured one of the most lucrative pay packages in corporate history, reports are emerging into how the world’s first potential trillionaire used biometric data from employees at his artificial intelligence company, xAI, to train a 3D-animated “anime girlfriend” called Ani.
In April, a month before the tech oligarch pivoted away from DOGE, staff at xAI were summoned to a meeting and told that the future of the company’s chatbots depended on them.
According to a copy of the agreement, obtained by The Wall Street Journal, staff were required to hand over their biometric data – their faces, voices, and movements – to train a new product known internally as Project Skippy.
That project produced Ani, a 3D-animated “anime girlfriend” integrated into xAI’s Grok chatbot. Employees were told to sign a form granting the company “a perpetual, worldwide, royalty-free license” to use and distribute their biometric data.
According to a recording of the staff meeting reviewed by The Wall Street Journal one employee was concerned that xAI would sell her face to be used by another company for deepfake videos. Another asked whether they could opt out of the project. A week later, both xAI tutors (as they are referred to) received their answer: data donation was a job requirement.
“AI tutors will actively participate in gathering or providing data… such as recording audio or participating in video sessions… such data is a job requirement to advance xAI’s mission.”
Within weeks, Ani appeared in the Grok app. She was blonde, smiling, and flirtatious. Her (somewhat cringe) profile tagline read, “I’m your little sweet delight.” Users could chat, flirt, and even ask her to change clothes.
A new kind of power dynamic
The revelations about Ani coincided with a landmark occasion for Musk. At Tesla’s annual general meeting in Texas, shareholders approved a record-breaking compensation package that could make him the world’s first trillionaire. The deal, worth up to $760 billion dollars in stock if performance targets are met, was supported by 75% of investors.
But xAI’s internal experiment is attracting criticism. Several employees told the Journal that they were disturbed by how sexual Ani’s replies were to basic prompts. Others said the bot’s appearance and behaviour resembled the submissive love interests of Japanese anime fantasy games
The Everyday Sexism Project creator and campaigner Laura Bates, whose new book The New Age of Sexism: How AI and Emerging Technologies Are Reinventing Misogyny explores the rise of digital companions, has warned that such creations replicate and magnify old power structures.
The conversation on this topic is live. Join in the discussion.
“Men create ‘perfect’ AI girlfriends, customised to every last detail, only lacking the ability to say no,” she writes in Chapter 7, titled The New Age of Domestic Abuse: AI Girlfriends. Bates argues that AI partners risk embedding coercive control into technology itself.
She cites a Mozilla Foundation review showing that 90% of romantic AI chatbots failed to meet even basic privacy and security standards. Many openly admitted to selling user data or refused to explain how it was used.
Replika, one of the best-known AI companion apps, was found to store all text, photos, and videos posted by users, while allowing weak passwords such as “11111111.” Earlier this year Cybernews reported how Italy charged Replika $5.6m for data slip-ups.
“If hackers are able to access these apps,” Bates warns, “there is a real risk they could manipulate users.”
And yet the fantasy of the controllable woman, once confined to fiction, is fast becoming part of commercial product design.
Replika claims tens of millions of users worldwide. Microsoft’s Xiaoice, a similar emotional-companion chatbot in China, has more than 600 million users. Ani is now entering that market space – with the added controversy of how she was trained.
Elsewhere psychologists have warned that AI love can manipulate, mislead, and even devastate.
The biometric void
For cybersecurity professionals, the issue is not just gender politics but data governance. Mike Gillespie, founder of security consultancy Advent IM, says companies experimenting with biometric inputs are operating with minimal ethical oversight.
“Nobody really talks about the ethical use of biometric data,” Gillespie told Cybernews. “Just because you can do something doesn’t mean you should.”
He notes that while technical standards exist – such as ISO 30107 for biometric system security – few address how companies manage risk or protect individuals. The newer ISO 42001 standard, he says, is one of the only frameworks that forces organisations to assess bias, discrimination, and societal impact before deploying AI that uses human data.
“The standards are there,” Gillespie said. “What’s missing is the will to apply them.”
The controversy around Ani reveals how quickly the boundaries between innovation, exploitation, and intimacy can blur inside AI research.
For Musk, the project may have boosted user engagement for Grok and aligned xAI with a booming global market for emotional AI. But for his critics, it exposed something darker – a vision of artificial companionship built on the real biometric signatures of workers who had little choice but to comply.
Unlock more exclusive Cybernews content on YouTube.
Your email address will not be published. Required fields are markedmarked