Employees at Elon Musk’s artificial intelligence startup xAI reportedly had to sign away the rights to their own faces and voices to help train the company’s next generation of chatbots — including a sexually suggestive virtual companion named “Ani.”
The demand, part of a confidential initiative called “Project Skippy,” required workers to grant xAI “a perpetual, worldwide, non-exclusive, sub-licensable, royalty-free license” to use, reproduce and distribute their biometric data, according to internal documents reviewed by the Wall Street Journal.
Most of the affected employees were so-called “AI tutors,” staff who work on the large language models that power xAI’s flagship chatbot, Grok.
At an April meeting led by company lawyer Lily Lim, employees were told xAI needed authentic human images and audio to make its digital avatars “act and appear like human beings,” The Journal reported.
On a recording of the session reviewed by the newspaper, one worker asked whether xAI could later sell their likeness to others.
Another employee pressed Lim to confirm if there was any option to decline participation.
“Could you just explicitly, for the record, let us know if there’s some option to opt out?” the person asked.
The project leader offered no such assurance, The Journal reported.
“If you have any concerns with regards to the project,” the leader was quoted as saying, “you’re welcome to reach out to any of the points of contact listed on the second slide.”
A week later, tutors received a notice titled “AI Tutor’s Role in Advancing xAI’s Mission,” informing them that recording audio or video sessions was “a job requirement.”
Some employees whose likenesses were used to train the avatars told The Journal they were disturbed by how sexualized “Ani’s” responses became.
Others worried their faces could be repurposed in deepfake videos or used without consent in other products.
Musk, who personally directed the creation of “Ani,” has defended the chatbots as tools for emotional connection.
“I predict — counter-intuitively — that it will increase the birth rate! Mark my words,” he wrote in August on X, the social platform he owns.
A New York Times report last month said “Ani” and her male counterpart, “Valentine,” were marketed as “sexy AI companions” and that Musk has been urging users to try them, even posting clips of the female bot dancing in lingerie.
Regulators are taking notice.
In August, 44 state attorneys general sent letters to xAI, Meta and other firms warning them to protect minors from explicit AI content. Meta reportedly changed instructions to its AI bot after leaked documents showed they were permitting so-called “sensual” chats.
Inside xAI, however, the focus remained on getting results fast, The Journal reported.
Former executives told the publication that Musk scrapped all-hands meetings and began personally overseeing Grok’s code, often holding sessions late into the night.
He wanted Grok — which competes with OpenAI’s ChatGPT — to become the world’s most popular chatbot.
That push came as Tesla, where Musk is CEO, has been facing new challenges.
Vehicle sales plunged 13.5% last quarter, marking the company’s second consecutive decline. Several major shareholders have questioned how much of his time he actually spends on the electric-car business.
Tesla board chair Robyn Denholm, who has advocated for an estimated $1 trillion pay package for Musk, has brushed off concerns, saying, “Other CEOs might like to play golf. He doesn’t play golf. So he likes to create companies, and they’re not necessarily Tesla companies.”
Tesla’s proxy filings appear to show Musk’s shifting priorities.
The company’s September report mentioned xAI 47 times, and shareholders are set to vote this week on whether Tesla should invest directly in Musk’s AI firm. Musk has supported the move.
The Post has sought comment from Musk, Denholm, Tesla and xAI.
