Online creators are creating deepfake memes as the technology grows more accessible, and publishing YouTube instructions so others can learn how
Karen Hao / MIT Technology Review : Tweets: @allanwhite , @techreview , @hkanji , @_karenhao , and @martinwaxman Tweets: Allan White / @allanwhite : Manipulation of photos, audio, & video is going to be rampant, basically from here on out. We need better tech tools to detect #deepfakes - but more importantly, a critical eye that doesn't blindly amplify or RT “hot” content. Slow it down. Triple check. https://twitter.com/... https://twitter.com/... @techreview : Easy-to-make deepfake memes are showing up on Twitter, Instagram, and especially TikTok. So our photo editor decided to make one of one of our AI writers. The rapidly increasing accessibility of this technology raises new concerns about its abuse. https://www.technologyreview.com/ ... Hussein Kanji / @hkanji : “There's a fine line between using deepfakes for entertainment and memes, and using them for harm” https://www.technologyreview.com/ ... Karen Hao / @_karenhao : Y'all I wrote a story about how easy it is to make deepfake memes and our photo editor @sparrow47 one-upped me by MAKING a deepfake meme of me for the cover art (bottom right). It's absolutely terrifying and hilarious at the same time, I'm deaddd. https://www.technologyreview.com/ ... https://twitter.com/... Martin Waxman / @martinwaxman : Deepfake videos are getting easier to create. Watch Grace Windheim explain how to make one step by step. @_KarenHao describes the process and output, which is easy to spot now, but likely won't be in the future https://www.technologyreview.com/ ... cc @alexsevigny @TristanLamonica #AIinPR
Context & Ripple Effects
The barrier to making a deepfake has collapsed in under two years: where one author documented a $552, two-week Faceswap effort in late 2019, MIT Technology Review's own photo editor could now produce a meme quickly enough that easy versions are spreading across Twitter, Instagram and especially TikTok — and creators are uploading YouTube tutorials so others can replicate them.
That puts this story at the hinge between two threads already in the corpus: AI firms like Synthesia building corporate-friendly synthetic video, and researchers warning since early 2019 that convincing fakes give cover for dismissing real events as fake, which is why DARPA funds detection work.
First-order effects
- TikTok, Twitter and Instagram now host a class of manipulated media made casually for laughs rather than deception, forcing their moderation systems to police content whose intent is ambiguous.
- Viewers encountering these memes face the exact problem Allan White flags in his tweet — manipulation is becoming rampant, so audiences must slow down and verify 'hot' content before amplifying it.
Second-order effects
- Demand for detection tooling intensifies, reinforcing DARPA-backed research efforts and pushing platforms toward provenance and verification features rather than manual review alone.
- Legitimate vendors like Synthesia, whose multilingual training videos position deepfakes as enterprise-safe, now share a technology stack with viral memes, so corporate buyers must weigh reputational contamination when adopting synthetic media.
Third-order effects
- If tutorial-driven diffusion continues, the trajectory runs toward what later coverage describes: tools like Stable Diffusion and DreamBooth letting anyone produce life-wrecking fakes from a few social-media photos, moving the problem from memes to targeted harm.
- Structurally, the 'liar's dividend' identified in 2019 hardens into default behavior — as fabricated media becomes ordinary entertainment, genuine footage becomes easier to dismiss as fake, degrading shared evidentiary ground for public discourse.
The trend: Synthetic media is diffusing from specialist projects to mass-produced casual content faster than detection and platform policy can keep up, shifting deepfakes from an information-war concern to a routine consumer phenomenon.