Skip to content
Submit incident
AII-1002-56713
ConfirmedA court, a regulator or the organization itself has stated the core facts, or the record cites a court or regulator source.Supporting sources are pending verification.

Grok generates photorealistic racist images of football players

Grok's new photorealistic image generation feature reportedly enabled users to create racist images of football players and managers starting in December 2024. Signify reported a sharp increase in online abuse complaints, and experts warned the problem will worsen as photorealistic synthetic media becomes more prevalent.

December 2024 Β· xAI Β· Harmful and abusive content
December 2024
AII-1002-56713
Quick facts
Category
Harmful and abusive content
Country
United Kingdom
Date
December 2024
Drafted by
Raidu Incident Detection

What happened

Signify, an organisation that tracks online hate in sports, reported seeing a sharp increase in racist image reports since Grok strengthened its image generation capabilities in December 2024. Examples include an image depicting a Black player picking cotton, another showing the same player eating a banana surrounded by monkeys in a forest, and images of players and managers depicted meeting controversial historical figures such as Adolf Hitler, Saddam Hussein and Osama bin Laden.

The new Aurora text-to-image feature created photorealistic AI images from simple user prompts. Users circumvent Grok's guidelines by 'jailbreaking', including describing physical features rather than naming individuals.

Reported impact

Athletes targeted by these images face personal harm and broader implications for their mental health and public perception. Signify warned that photorealistic AI will make such outputs far more prevalent and that the normalisation of this imagery can contribute to a toxic environment in sports.

The Premier League received more than 1,500 racist abuse reports in 2024 and has a dedicated team to find and report racist abuse directed towards athletes.

Index analysis

Unlike other AI tools with stricter guidelines, Grok's system has been shown to be easily manipulated or jailbroken by users to generate inappropriate images.

Classification

Category
Harmful and abusive content
Industry
Media, social platforms and marketing
Country
United Kingdom
Harm status
Actual harm
Intent
Intentional
Timing
Post-deployment
Harm types
Human or fundamental rights, Psychological, Reputational
Failure modes
Deepfake image, Jailbreak, Manipulation or dark pattern, Content moderation failure
AI systems
Grok

Sources and evidence

Sources not yet published

This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.

How sources are verifiedSuggest a source for this incident