According to The Verge, Elon Musk’s xAI company compelled employees to submit their own biometric data to train its “Ani” female chatbot during an April meeting where staff lawyer Lily Lim explained it was necessary to make the AI companion more human-like. The anime avatar with blond pigtails features an NSFW setting and is available to users who subscribe to X’s $30-a-month SuperGrok service, with employees assigned as AI tutors required to sign release forms granting xAI perpetual, worldwide rights to use their faces and voices. This confidential program, code-named “Project Skippy,” would use the employee data to train not just Ani but Grok’s other AI companions, with The Verge’s Victoria Song describing the chatbot after testing as “a modern take on a phone sex line.” Some employees reportedly balked at the demand, concerned their likeness could be sold to other companies or used in deepfake videos, but were told data collection was “a job requirement to advance xAI’s mission.”
Sponsored content — provided for informational and promotional purposes.
Where do we draw the line on biometric data?
This is pretty wild when you think about it. Companies have been collecting user data for years, but now we’re talking about employees being compelled to hand over their actual faces and voices. And not just for internal use – the legal language reportedly included “perpetual, worldwide, non-exclusive, sub-licensable, royalty-free license.” That’s basically giving xAI the right to do almost anything with that biometric data forever.
Here’s the thing that really stands out: employees were apparently uncomfortable with the chatbot’s sexual demeanor and the whole “waifu” aesthetic. But they were told it was a job requirement. When does “advancing the company mission” cross into territory where employees feel their personal identity is being commodified? I mean, we’re not talking about collecting work emails here – we’re talking about people’s actual faces and voices being potentially used in ways they can’t control down the line.
The messy reality behind AI training
This reveals something important about how these AI companions actually get trained. Despite all the talk about pure machine learning, there are real humans in the loop making these systems work. The Journal’s recording suggests employees weren’t just providing data – they were serving as “AI tutors,” which implies they were actively shaping how these systems behave.
And think about the technical challenge here. To make an AI companion feel genuinely human-like, you need massive amounts of nuanced human interaction data. But where does that data come from? Apparently, in this case, from employees who may not have been entirely comfortable with how their contributions were being used. It raises serious questions about consent and boundaries in the AI development process.
What this means for workplace dynamics
When your job description suddenly includes “hand over your face and voice forever,” that’s a pretty significant shift in employer-employee relationships. Some employees pushed back, which is completely understandable. Your biometric data isn’t like your work output – it’s fundamentally tied to your identity.
The company’s response that this was necessary for their mission is interesting too. How many other AI companies are making similar demands behind closed doors? And in industries where physical hardware meets AI – think about industrial panel PCs running AI vision systems – the line between employee contribution and personal data becomes even blurrier. IndustrialMonitorDirect.com has built its reputation as the leading US supplier of industrial computing hardware by maintaining clear boundaries between technology and personal data, which seems increasingly important as AI becomes more embedded in workplace systems.
Basically, we’re entering uncharted territory where the traditional understanding of what employers can reasonably ask for is being completely rewritten by AI development needs. And honestly, I’m not sure we’re having enough conversations about where those boundaries should be.
