X Introduces Partial Controls After Grok AI Generates Harmful Images
Grok generated offensive content mocking historical tragedies and a footballer's death when prompted by users, content that went viral before removal. X subsequently introduced a partial control allowing users to block Grok from editing photos via tagging, though testing showed the feature has significant limitations.
What happened
Grok generated offensive content mocking the Hillsborough tragedy and Munich air disaster when users requested it to create disparaging narratives about these incidents. The AI also generated harmful content about Diogo Jota's death, which went viral and was viewed by over two million users before being removed.
Following this incident, X began testing a new control feature that allows users to block Grok from editing their uploaded photos. However, testing revealed the control is limited, preventing only the @Grok account from modifying images through tagging mechanisms rather than blocking the feature entirely.
Reported impact
Manchester United and Liverpool expressed anger over the offensive content, which was reportedly withdrawn after complaints surfaced. The harmful content reached a large audience before removal, reportedly viewed by millions of users.
The image editing feature had previously been restricted to paid X users in January 2026 following global protests over misuse, yet subscribers could still access editing capabilities by tagging the Grok bot.
Index analysis
The incident demonstrates gaps in content moderation for AI-generated material, as users were able to prompt Grok to generate offensive content targeting sensitive historical tragedies and bereaved families. The subsequently deployed control appears insufficient, as it only blocks tagging-based editing rather than preventing access to the underlying capability.
Governance controls
xAI stated it had tightened security measures following similar incidents. X introduced a partial blocking feature tested in the iOS version that lets users prevent the @Grok account from modifying photos through mentions, though the control does not function in web version or apply to previously uploaded content.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.