OPEN LINK ↗
// 1h agoSECURITY INCIDENT
Roblox Crawler Scrapes 145GB, Including Adult Models
Open3DLab says a crawler identifying itself with a Roblox email downloaded 1,027 files totaling 145.5GB from 712 projects, including 345 adult projects and 423 marked non-commercial. Roblox has not confirmed operating the crawler or using the files for AI training. [Source](https://open3dlab.com/news/roblox-downloaded-145-gb-of-models-from-our-sites/)
// ANALYSIS
This is a data-governance failure before it is a proven AI-training scandal: the logs are detailed, but attribution and downstream use remain unverified.
- –The `trellis2-data-sourcing` label aligns with Roblox’s public Cube 3D foundation-model work, making the allegation plausible but not conclusive. [Roblox Cube](https://about.roblox.com/newsroom/2025/03/introducing-roblox-cube)
- –Ignoring robots.txt, terms of service, and creator license restrictions turns a free-download corpus into a consent and provenance problem.
- –Adult assets were reportedly fetched at the same rate as everything else, exposing weak filtering for a platform heavily used by children.
- –AI developers need operational provenance controls: allowlists, opt-outs, license checks, sensitive-content quarantine, and deletion propagation.
// TAGS
robloxsecuritydatasettraining3d-genethics
DISCOVERED
1h ago
2026-10-27
PUBLISHED
3h ago
2026-10-27
RELEVANCE
8/ 10
AUTHOR
choult