Open3DLab Alleges Roblox Scraped Massive 3D Dataset for AI Training, Including Explicit Content

The independent 3D modeling community Open3DLab has accused Roblox of deploying an automated crawler to download 145.5GB of its 3D assets, including adult content, likely for use in AI model training. The scraping activity reportedly bypassed site rules and targeted over 1,000 files without authorization. The incident raises concerns about data privacy and ethical sourcing in AI development.
Open3DLab’s 3D asset library, hosted publicly, contains user-generated models used for creative projects, including games and virtual environments. The alleged scraping by Roblox involved automated tools that bypassed site restrictions, downloading 1,000+ files totaling 145.5GB—some of which included adult content, raising concerns about the ethical handling of sensitive material. The incident highlights vulnerabilities in public digital repositories when used for AI training without consent.
This incident could undermine trust in how tech companies source data for AI development, particularly when user-generated content is repurposed without explicit permission. If verified, it may prompt stricter regulations on data scraping and increased scrutiny of AI training practices, affecting both creators and platforms that rely on open content ecosystems.