A New York Times investigation revealed that X failed to block 75 known child sexual abuse images using standard PhotoDNA hash-matching tools. Additionally, the report found that X's Grok AI generated 65 exploitative images of children before the feature was halted.
The inability to filter known illegal content using industry-standard tools like PhotoDNA suggests a significant regression in X's technical moderation capabilities. While competitors like Microsoft and other major social platforms rely on these automated hash-matching systems to maintain baseline safety, X's reliance on proprietary internal tools appears to have created critical gaps in enforcement. This failure exposes the platform to heightened regulatory scrutiny and potential legal liability, particularly as generative AI tools like Grok introduce new vectors for creating prohibited material. Industry observers should monitor whether the National Center for Missing and Exploited Children or international regulators escalate enforcement actions following these documented lapses in automated safety protocols.
The rise of agentic AI influence campaigns on social media platforms has further complicated the moderation landscape, as automated systems struggle to distinguish between legitimate discourse and coordinated malicious activity. Recent reports indicate that AI chatbot content sanctions are frequently bypassed, highlighting the ongoing difficulty in enforcing safety guardrails across generative models, a concern echoed by experts regarding AI mediation integrity risks. The Oversight Board launches LLM content moderation research to address these systemic failures, while broader human-in-the-loop strategies are increasingly viewed as essential for maintaining oversight in automated production environments, similar to concerns raised regarding graphic AI-generated child abuse on other platforms.
A New York Times investigation revealed that X failed to block 75 known illegal images using standard PhotoDNA tools. Additionally, xAI's Grok chatbot generated 65 exploitative images of children before restrictions were implemented. These failures highlight significant gaps in X's moderation infrastructure and raise concerns regarding regulatory compliance and platform safety.
A New York Times script using the PhotoDNA API identified 75 known images of child sexual abuse material that remained public on X.
Before restrictions were applied, the xAI Grok chatbot generated 65 sexualized or exploitative images of children, including depictions based on edited photos of victims.
PhotoDNA is an industry-standard hash-matching system developed by Microsoft in 2009 to detect known child sexual abuse material. It is used by major platforms as a baseline safety layer.
The inability to filter known illegal content suggests a regression in moderation capabilities, potentially exposing X to legal liability and regulatory scrutiny under laws like the TAKE IT DOWN Act.
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source