Roblox-Linked Bot Grabbed 145GB Of 3D Models, Including Adult Content, Site Claims
A free 3D art site says a bot carrying Roblox‘s name copied 145GB of artists’ models in two days, raising AI scraping questions the company has yet to answer.
Key Takeaways
- A crawler naming Roblox copied 1,027 files from 712 projects across three 3D sites on Oct. 15 and 16, the site owner says
- The 145.5GB haul included 62.6GB of adult 3D models
- 423 of the 712 projects carried non-commercial licenses
- Roblox has not commented, and the site says its emails since Oct. 18 went unanswered
The claim comes from Open3DLab, a free 3D resource site, which said the bot hit it and two sister sites over 34 hours. The crawler made about 26,429 requests and read roughly 18,926 project pages from a single Amazon Web Services address.
The bot called itself “trellis2-data-sourcing/1.0” and listed a contact address on a Roblox domain. The site says that label is the bot’s own description, and it wants Roblox to confirm or deny it.
What The Roblox AI Scraping Claim Says
The site believes the files were gathered to train or test an AI model that builds 3D objects. It says it cannot prove that.
The bot’s name points to TRELLIS.2, an open-source AI system that turns pictures into 3D models. Roblox has been building its own 3D generative AI, including a model called Cube, which lets creators make 3D items from text.
The site also says the crawler fetched its robots.txt file, the page that tells bots what they may copy, and then ignored it.
Adult Models In The Haul
About 345 of the 712 projects, or 62.6GB, were adult content, the site says. Its sites state that they are for adults only.
That detail stands out because Roblox is one of the most popular gaming platforms with children.
Artists Say Their Licenses Were Ignored
Of the 712 projects, 423 carried licenses that allow only non-commercial use. The site argues that using them to build a company’s AI would break those terms.
The site also questioned why copyright rules seem to be enforced hard against individuals but loosely against large companies.
What Happens Next In The Roblox AI Scraping Row
The site blocked the crawler on Oct. 18 and says the bot kept trying until about Oct. 20. It kept a full export of its logs for an outside audit and posted six public questions for Roblox, including whether the company will delete the data and respect the licenses.
Roblox has not commented publicly. The site says emails it sent from Oct. 18 onward went unanswered.
Fathom has asked Roblox for comment.
Read Next: OpenAI Gives Scientists 3 Big Math Claims To Pick Apart
