Open3DLab, a community hub for 3D assets, reported that a crawler identified as belonging to Roblox systematically scraped its infrastructure between October 15 and 16, 2026. The crawler, which operated from a single Amazon Web Services (AWS) address, accessed nearly the entire catalog of the site, eventually downloading 1,027 files totaling 145.5 GB. The crawler used a specific user agent—`trellis2-data-sourcing/1.0 (3D foundation model research; [redacted]@roblox.com)`—and ignored the site's `robots.txt` file, Terms of Service, and individual creator licenses, many of which were restricted to non-commercial use.
The data collected included high-quality, rigged 3D characters and assets that are often scarce in public datasets. Notably, the crawler did not distinguish between general content and adult material; it downloaded 364 files (54.8 GB) from SmutBase, a site exclusively for adult content, and 345 projects (62.6 GB) marked as adult on SFMLab. Open3DLab notes that the crawler's user agent is linked to a researcher who is one of the authors of Roblox's 'Cube' project, a generative AI system for 3D and 4D intelligence. While the site operators have blocked the crawler and contacted Roblox for comment, the company has not yet responded to their inquiries.
The incident highlights a significant concern regarding the safety of children on the Roblox platform, where over a third of daily users are under 13. Because the crawler did not filter for adult content during the collection phase, there is a risk that explicit data could influence the underlying foundation model, even if filters are applied later. Furthermore, the report underscores a growing disparity in copyright enforcement. The author argues that while individual creators face severe legal consequences for minor infractions, large tech companies can treat massive, unauthorized data scraping as a negotiable cost of doing business, creating a 'two-tier system' of copyright protection.