Alibaba's paper on its Animate Anyone AI model goes viral after videos show the model can replace TikTok influencers after being trained on their videos
the thing it was inadvertently made to accomplish—is the destruction of the internet, by making it so trivially easy to falsify anything that the internet ceases to reliably serve any of the purposes for which we use it, so that eventually we either find some other reason to use the internet, or we stop using it altogether. Bluesky: Mary Branscombe / @marypcbuk.bsky.social : actually, I feel like I'm OK with AI replacing anyone called an ‘influencer’ on any platform because it's not like it will be displacing quality content... [embedded post] X: Samantha Cole / @samleecole : if I had to guess, I'd say Alibaba is working on an app or something where you can choose to insert yourself or someone else into a library of popular dances. will those choreographies be credited or compensated? reminds me of this: https://www.npr.org/... Samantha Cole / @samleecole : this, for example, is Jasmine Chiswell, a hugely popular youtuber and tiktoker. wanna take bets on whether she approved of her content being turned into AI dog food by Alibaba, so people could say how excited they are for her to lose her job? https://twitter.com/... Samantha Cole / @samleecole : regardless of what you think of tiktokers, companies like Alibaba are playing a legally risky game. last month, a federal judge said “reducing choreography to poses would be akin to reducing music to just notes,” giving artists the go-ahead to keep suing https://www.404media.co/... [image] Samantha Cole / @samleecole : this week, a new AI technique for making dance videos had everyone saying “RIP TikTokers.” the irony is, stealing from content creators is blatantly baked in: https://www.404media.co/... @dreamingtulpa : RIP TikTokers 💀 [video]
Context & Ripple Effects
Short-form video already depends on a retention-oriented feed system: TikTok’s internal engineering documents described optimization around retention and time spent in TikTok’s engagement-driven recommendation system. Animate Anyone puts identity-preserving synthetic performance closer to that distribution loop.
The immediate controversy is not just about generated clips but about whether creators’ videos can become training material or a substitute for the creator. That tension later surfaced in reporting that TikTok was testing prompt-generated virtual influencers for advertising.
First-order effects
- Creators whose videos are used or imitated face a more immediate risk of unauthorized likeness use and of their recognizable performance being reproduced without their participation.
- Alibaba’s model demonstration makes synthetic influencer-style video a concrete product and reputational issue, intensifying scrutiny of the provenance and consent behind training examples.
Second-order effects
- TikTok, advertisers, and creator-management businesses gain an incentive to test cheaper virtual talent while developing policies for disclosure, rights clearance, and impersonation complaints.
- The demonstrated workflow strengthens the commercial value of licenses covering a performer’s image, movement, and video archive—similar to the later talks over licensing studio content for AI video tools.
Third-order effects
- If creator video increasingly functions as inference input, the creator economy could shift from paying for each new performance toward negotiating rights over reusable identity and training data.
- The key competitive question may become which platforms can pair generative tools with trusted provenance, consent, and distribution controls; absent those controls, synthetic supply can weaken confidence in what viewers see.
The trend: This is one early instance of AI content commercialization turning social-video creators from content suppliers into potential training-data and likeness-rights holders.