AI-Generated Black Female Avatars Spark Racism Concerns Across Social Media

TikTok has deleted 20 accounts after a BBC investigation that revealed a troubling trend of AI-generated black female avatars deployed to drive users towards adult content. The platform acted after the BBC and experts at the AI research outlet Riddance discovered dozens of accounts across Instagram and TikTok featuring highly sexualised digital characters with exaggerated proportions and darkened skin. The accounts, which were unlabelled as AI-generated in clear violation of platform guidelines, employed racist tropes in their naming conventions and promotional language. Whilst TikTok has acted swiftly, Meta, Instagram’s parent company, said it was looking into but has failed to confirm taking similar action against the accounts active on its platform.

The Discovery: Numerous Misleading User Accounts

The BBC’s investigation, carried out alongside Riddance researchers Jeremy Carrasco and Angel Nulani, uncovered a complex scheme operating across numerous platforms. The team discovered 60 accounts mainly found on Instagram that included links directing pointing users toward paid sexually explicit content on third-party websites. Notably, whilst these third-party sites labelled the imagery as AI-generated, the Instagram accounts themselves provided no such information, resulting in calculated dishonesty for unsuspecting users. The research revealed an even larger ecosystem of comparable profiles on Instagram and TikTok alike that contained no links to commercial material, suggesting the problem stretches considerably further than profit-driven abuse.

The accounts utilised a deliberate approach to evade detection and grow audiences at pace. Nearly all accounts were based primarily on Instagram, with around one in three also maintaining versions on TikTok. Account names specifically drew upon racially charged language, including terms such as “black”, “noir”, “dark” and “ebony”, alongside language promoting racial stereotypes and objectification. Many accounts followed or engaged with each other, creating an interconnected network that expanded their prominence and visibility. This organised strategy points to intentional organisation rather than standalone occurrences, raising serious questions about the scope and technical capability of the operation.

• AI avatars displayed distorted physical proportions and darkened skin tones
• Account names employed racially charged language and stereotyped characterisations about white men
• Videos lacked labels as generated by artificial intelligence, violating platform guidelines
• Many accounts were interconnected, establishing coordinated systems for amplification

Abuse Through Artificial Creation

Stolen Content and Digital Manipulation

The investigation revealed a especially concerning dimension to the enterprise: the wholesale theft of content from authentic content creators. One channel that gathered three million followers within weeks of its launch in December had deliberately repurposed content from genuine individuals, most notably Malaysian model Riya Ulan. The perpetrators overlaid the AI-generated avatar’s face—featuring an darkened skin tone—onto Riya’s body, precisely duplicating her gestures, outfits and background settings. This brazen act of digital theft intensified the original exploitation, converting genuine creative work into fuel for misleading accounts.

Riya’s story underscored the breach of privacy inherent in such methods. Upon discovering her content had been stolen and repurposed, she voiced her upset to the BBC: “I was angry. Of course my videos are widely distributed… It doesn’t mean that you can simply steal it and use it as your own.” Her distress exposes a critical gap in safeguarding systems, where creators lack sufficient defences against their identities being misused for fake profiles. The incident shows how AI technology can amplify current abuses, permitting fraudsters to expand their abuse across numerous creators in parallel.

The exploitation extended beyond simple video theft. Accounts intentionally created synthetic characters using exaggerated bodily features and artificially manipulated skin tones to produce eye-catching, visually arresting material crafted to generate interaction. These synthetic entities were marketed as authentic influencers, accompanied by false narratives and personal narratives, at the same time perpetuating harmful racial stereotypes and sexualised stereotypes. The sophistication of the deception caused numerous people interacted with the posts believing they were engaging with genuine individuals, not algorithmic constructs created expressly to drive traffic toward harmful commercial offerings.

• Avatar’s face overlaid onto pirated content from authentic creators
• Artificially darkened complexions and pronounced characteristics designed for engagement
• Accounts positioned as genuine influencers with false backgrounds and identities

Racist Depictions and Harmful Stereotyping

The accounts uncovered by the BBC and Riddance researchers showed a troubling pattern of discriminatory targeting through their intentional account names and messaging. Account identifiers incorporated terms such as “black”, “noir”, “dark” and “ebony”, whilst posts repeatedly featured race-based terminology and fetishizing content. Many included statements such as “loves white men” and “why I need a white guy in my life”, sustaining racist tropes and reducing Black women to objects of sexual desire. This linguistic framework rendered the profiles into stereotypical representations, perpetuating long-established racist tropes that commodify and dehumanise Black femininity.

The graphic depiction intensified these language-based injuries. The avatars were regularly shown in revealing swimwear and skimpy clothing, their bodies algorithmically altered into exaggerated proportions designed to maximise interaction via sexual imagery. Complexions were synthetically deepened to create an artificial, near monstrous appearance that bore little resemblance to authentic human variation. This mix of distorted characteristics, exposed clothing and synthetically altered characteristics produced algorithmic stereotypes that operated like modern-day minstrelsy, converting racist imagery into algorithmic content designed for widespread distribution and monetisation.

Why AI Magnifies the Problem

Artificial intelligence technology has significantly changed the scale and sophistication of racist exploitation online. Previously, such harmful content required significant resources and manual effort to produce; AI generation enables widespread production of racist caricatures, enabling bad actors to produce numerous realistic fake accounts at minimal cost. The technology’s ability to create photorealistic imagery imparts unwarranted credibility to racist tropes, making them appear credible to unsuspecting users. This technological advantage turns what was once fringe exploitation into a expandable, revenue-generating enterprise.

The computational nature of social media platforms amplifies these harms dramatically. AI-generated content tailored for user interaction—particularly content that exploits racial and sexual stereotypes—spreads rapidly through recommendation systems built to maximise user interaction. Platforms find it difficult to manage the vast quantity of AI-generated content, whilst the absence of required transparency markers means users cannot distinguish authentic creators from computational entities. This creates an environment where racist stereotypes proliferate unchecked, laundered through the veneer of technological innovation and presented as entertainment rather than exploitation.

Platform Accountability and Response

TikTok’s rapid action in deleting 20 accounts demonstrates a infrequent example of platform action, yet it underscores the reactive approach of content moderation in the social media age. The bans came solely following sustained media scrutiny and public pressure, suggesting that in the absence of external investigation, these accounts would have remained active unchecked. Notably, TikTok’s response was confined to a single platform, whilst numerous problematic accounts sustained active accounts on Instagram, where they went on gathering followers and routing traffic to harmful third-party platforms. This disjointed response underscores the insufficiency of individual platform responses to widespread challenges.

Meta’s claim to be investigating, combined with its refusal to communicate specific steps, exposes the disconnect between corporate accountability rhetoric and actual implementation. The company’s sluggish reply contrasts sharply with the critical imperative to combat swiftly spreading damaging material. Both platforms have developed policies mandating disclosure of machine-generated visual content, yet these standards stayed unimplemented across numerous profiles. This implementation shortfall indicates that without legislative action and juridical responsibility, platforms will persist in prioritizing user engagement and ad income over platform security and the safeguarding of vulnerable groups from organised harm and bigoted material.

Broader Implications for Digital Integrity

The rapid growth of unidentified AI-generated material on leading social networks raises key questions about online integrity and user trust. As AI technology becomes increasingly sophisticated, telling apart real content creators and synthetic avatars grows exponentially more difficult for ordinary users. This loss of genuineness undermines the core assumption upon which social media communities are built—the assumption that accounts represent actual individuals sharing genuine experiences. When millions of users interact without realising with artificial identities designed to exploit them, the networks’ trustworthiness as reliable platforms for interaction declines sharply. The lack of required transparency measures deliberately misleads users and compromises informed consent.

Beyond individual deception, the systematic use of artificially created black female avatars represents a particularly insidious exploitation of racial identity in digital spaces. These synthetic personas exploit harmful stereotypes and hypersexualised racial tropes whilst concurrently appropriating real creators’ content and labour. The practice reinforces the commodification of black femininity within algorithmic systems designed to maximise engagement and profit. This algorithmic discrimination operates at scale, reaching millions whilst remaining largely invisible to platform enforcement teams. Without robust regulatory structures requiring openness, verification protocols, and meaningful consequences for violations, digital spaces will keep enabling widespread damage against marginalised communities.

• AI disclosure requirements must be binding in law on every platform worldwide
• Authentication systems ought to validate the creator’s identity before monetisation or follower accumulation
• Racial discrimination within AI-generated content necessitates direct restrictions and ongoing monitoring
• Platforms must implement real-time moderation with penalties aligned to content seriousness