The rapid proliferation of generative artificial intelligence has injected a dangerous new potency into the old poison of racial caricature, escalating the phenomenon of “digital blackface” from online appropriation to a tool for mass disinformation and political propaganda. This modern minstrelsy, where non-Black individuals and entities assume Black identities and stereotypes through digital means, has seen a massive acceleration in the past two years as AI video tools have become widely accessible, according to experts.
Benefit Scams and Viral Lies
A stark illustration emerged late last year during a US government shutdown, which cut off SNAP benefits for low-income families. Social media flooded with AI-generated videos purporting to show Black women gleefully abusing the system. One viral TikTok featured a woman boasting about selling $2,000 worth of food stamps for cash, while others ranted about taxpayers funding large families or melted down at store counters. Visible but faint AI watermarks did little to stop conservative commentators and some news outlets from presenting the frenzy as real. Fox News reported on the clips as authentic before issuing a correction, while commentators like Amir Odom lamented the alleged abuse. Newsmax anchor Rob Schmitt claimed people used benefits “to get their nails done, to get their weaves and hair,” and has elsewhere questioned why “300 pound” individuals receive subsidies, calling the programme a “scam.” Lost in this outrage was a key fact: white Americans constitute 37% of SNAP’s 42 million beneficiaries, and programme fraud is estimated at just 1%, with the average monthly payment often insufficient for groceries.
These clips represent a weapons-grade shift in digital blackface, moving beyond blending in to actively perpetuating harmful stereotypes. The term itself, coined in a 2006 academic paper by Joshua Lumpkin Green and popularised by cultural critic Lauren Michele Jackson in 2017, describes the commodification of Black identity for non-Black expression online. Earlier forms included white gamers using Bitmojis of a different race and adopting African American Vernacular English (AAVE), or the use of darker-skinned emojis and reaction memes featuring Black celebrities. “The digital blackface videos are really pulling from the same racist and sexist stereotypes and tropes that have been used for centuries,” says Safiya Umoja Noble, a UCLA gender studies professor and author of Algorithms of Oppression.
AI Tools and Synthetic Stereotypes
The technology fuelling this escalation is readily available. OpenAI’s text-to-video app Sora, which surged in popularity in 2025, has been used to create hyperrealistic deepfakes that sully the image of civil rights icon Martin Luther King Jr., showing him shoplifting or swearing through his “I Have a Dream” speech—a practice critics dubbed “synthetic resurrection.” Other firms like Hume AI offer synthetic voices for podcasts and audiobooks, such as “Black woman with subtle Louisiana accent,” often scraping speech from online creators without compensation or consent. This divorces Black expression from authorship, context, or consequence, with AI models absorbing tone and slang from digital spaces pioneered by Black users.
This digital caricature has found a powerful amplifier in political spheres. The Trump White House has been accused of leveraging these tactics. In January, the official White House X account posted a doctored photo of Minnesota activist Nekima Levy Armstrong, darkened and weeping, altered from an actual Department of Homeland Security image. Earlier this month, an image portraying the Obamas as apes was circulated via Trump’s own Truth Social account, reviving a longstanding racist slur. Trump disclaimed direct responsibility for that post, which was later taken down.
Roots in Minstrelsy, Responses from Tech
The practice has deep historical roots in the minstrel revues of the early 19th century, where white performers used burnt cork to caricature Black features and perform exaggerated routines of laziness and buffoonery. The character “Jim Crow,” played by Manhattan playwright Thomas D Rice, became shorthand for segregation policies. Even as minstrelsy faded, its residue lingered in culture, and now finds new life in AI. “The early research that was done on digital blackface started with white gamers using bitmojis of a different race and changing their vernacular to represent themselves,” says Mia Moody, a Baylor University journalism professor and author of the forthcoming Blackface Memes. “That’s part of the cultural appropriation, gaining the cultural capital.”
Tech firms have made halting efforts to respond. Bowing to public backlash, OpenAI, Google, and Midjourney disallowed deepfakes of King and other icons. Meta deleted two of its own AI blackface characters—”Grandpa Brian” and “Liv”—after criticism over their non-diverse development team. Platforms like Instagram and TikTok attempt to scrub such videos, with limited success; last summer, an AI avatar called “Bigfoot Baddie”—a Black woman-yeti hybrid created by Google’s Veo AI—spawned a social media craze and remains online. Social media platforms have begun labelling AI-generated content more broadly.
Organisations like Black in AI, the Distributed AI Research Institute (DAIR), the AI Now Institute, and the Partnership on AI have pushed for greater diversity in AI model-building and highlighted risks like systems learning from marginalised communities’ data without consent. They advocate for mechanisms like data opt-outs to limit exploitative uses. However, widespread adoption of such safeguards has been glacial. “With AI generation, these tech firms cannot manage what’s coming through their systems,” Noble warns, noting that authoritarian regimes can use these tools to facilitate propaganda.
An Engine for Official Disinformation
The precise impact is hard to quantify, but its use by the Trump administration underscores its potential as a tool for state-backed disinformation. The altered image of Armstrong scanned as a psychological operation by a government tracking activists, scholars note. Noble connects this to a broader political climate: “We are living in a United States with an open, no-holds-barred, anti-civil-rights… agenda. Finding the material to support this position is just a matter of the state bending reality to fit its imperatives.”
Beyond laundering bigotry as news, digital blackface exposes Black users to personalised abuse echoing minstrelsy’s heyday. The ethical implications are severe, involving the erasure of cultural context, reinforcement of stereotypes, and exploitation of marginalised cultures. There is a growing call for AI systems to be developed with cultural sensitivity and inclusivity, moving beyond Western-centric datasets.
Despite the challenges, some see a cyclical pattern. Moody remains hopeful that the current fascination with digital blackface will wane. “Right now people are just experimenting with AI technology and having a ball seeing what they can get away with,” she says. “Once we get beyond that, then we’re going to see less of it. They’ll move on to something else.” Yet, as history shows, the tropes are enduring, and the technology to animate them is now in the hands of anyone with an internet connection.
