TL;DR
Get privacy and security gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Open3DLab says a crawler identifying itself as Roblox downloaded 1,027 files totaling 145.5GB from Open3DLab, SFMLab and SmutBase on Oct. 15–16, 2026. The site operators say they blocked later attempts and have not received a reply from Roblox; the files’ intended use and whether they entered a model remain unconfirmed.
Open3DLab says a crawler identifying itself as conducting 3D foundation model research downloaded 1,027 files totaling 145.5GB from Open3DLab, SFMLab and SmutBase on Oct. 15 and 16, including adult content. The site operators say they blocked further attempts and contacted Roblox, but have received no reply; they have not confirmed that the files were used to train a model.
In a report published Oct. 18, Open3DLab said it found the crawler in its logs after it had read pages for 18,926 projects, described as practically the full catalog, through 26,429 requests over 34 hours. The crawler used the same user-agent string on each request: “trellis2-data-sourcing/1.0 (3D foundation model research; [redacted]@roblox.com)” and came from a single Amazon Web Services address, according to the site.
Open3DLab said the downloads covered 712 projects. Of the files, 364, totaling 54.8GB, came from 224 projects on SmutBase, which the report describes as an adult-content site. Including projects marked adult on SFMLab, the operators counted 345 adult projects and 62.6GB of adult material among the downloads.
The report says the crawler retrieved the sites’ robots.txt file but did not follow its restrictions, and that the downloads conflicted with the sites’ terms, which prohibit scraping. Open3DLab also said 423 of the 712 projects were licensed for non-commercial use only. These are the site operators’ findings; the report does not include a response from Roblox or independent verification of the crawler’s ownership.
Creator Licenses and Adult Files
The report raises questions about how public-facing 3D assets are collected when creators may have granted only limited reuse rights. A file being accessible online does not, by itself, establish permission to scrape it or use it in commercial model development. Open3DLab says the crawler downloaded work from projects whose licenses restricted commercial use, but the full license terms, any permissions Roblox may have held, and the crawler’s intended use have not been independently established.
The presence of adult material also matters because the sites say the crawler collected it at a rate similar to other content in the sample. That does not show how Roblox might handle the files, or whether they will be used at all. It does make the reported collection relevant to questions about data filtering, consent and safeguards for training datasets, particularly if any collected material were later incorporated into systems used to generate 3D content.
For creators, the immediate concern is whether their models were copied and what remedies are available. Open3DLab says it blocked further requests, but its post does not say that the downloaded copies were deleted or identify an agreement with Roblox about their handling.
Products Worth Considering
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
DJI Osmo Mobile 8 Gimbal with Mic Mini Transmitter, Infinity Black
As an affiliate, we earn on qualifying purchases.
Roblox’s 3D Model Research
Open3DLab links its suspicion about the crawler’s purpose to Roblox’s public work on Cube, a generative AI system for 3D and 4D content. The report says Roblox introduced Cube in March 2025 and that a paper described a goal of developing a foundation model for 3D intelligence. According to Open3DLab, the system’s initial shape model was trained on roughly 1.5 million assets from licensed and public datasets, including Objaverse, as well as free Creator Store assets whose creators had opted in. Roblox began rolling out 4D generation built on Cube in February 2026, the report says.
Open3DLab also points to the crawler’s user-agent wording and an email address it says is associated with a researcher among the authors of the Cube paper. The site argues that collecting complete, textured and sometimes rigged character models could fit foundation-model research. It explicitly says it does not accuse that researcher of acting independently, and assigns responsibility for a collection run to the company. The connection to Cube, however, is the site’s inference, not a confirmed explanation from Roblox.
““145.5 GB of our creators’ work, downloaded for what we assume are AI training purposes.””
— Open3DLab
Purpose and Data Use Unconfirmed
Roblox has not replied to the two emails Open3DLab says it sent on Oct. 18 and 20. The company has not publicly confirmed in the material provided that the crawler belonged to Roblox, explained why it accessed the sites, or said whether it had permission to collect any of the files.
Open3DLab says the crawler’s label and the company’s known 3D research make training or evaluation data collection the most likely explanation. That remains an inference. The report does not establish whether any files were used for training, evaluation, testing or another purpose, whether copies were retained, or whether they contributed to a model. It also does not provide the detailed log analysis the operators say they plan to publish.
The report gives its own account of the crawler’s identification, volume and licensing conflicts. No independent technical audit or response from the affected creators is included in the supplied material, so the scope of creator permissions and any direct impact remain unclear.
Further Logs and Roblox Reply
Open3DLab says it plans to publish a more detailed account of its logs. That could clarify the sequence and scale of the requests, and provide additional evidence about what the crawler accessed. The site says it blocked the later attempts: activity resumed on Oct. 19, with requests blocked, and the last reported return began at 8 a.m. on Oct. 20 before stopping around 3 p.m. Central European Summer Time.
The next key development would be a response from Roblox addressing the crawler’s ownership, purpose, authorization and handling of any downloaded material. Until then, the central question—whether the files were used in model development—remains unanswered. Open3DLab has not reported an agreement to delete the files or a process for notifying affected creators.
Key Questions
What does Open3DLab say the crawler downloaded?
It says the crawler took 1,027 files totaling 145.5GB from 712 projects across Open3DLab, SFMLab and SmutBase on Oct. 15 and 16, 2026.
How much of the reported material was adult content?
Open3DLab counted 364 files totaling 54.8GB from 224 SmutBase projects. Including adult-marked SFMLab projects, it said 345 projects and 62.6GB of the downloaded material were adult content.
Has Roblox confirmed that it collected the files for AI training?
No. Open3DLab suspects the collection was related to 3D foundation-model research, but says it cannot prove the purpose or whether files entered a model. The supplied report says Roblox has not responded to its emails.
Did Open3DLab stop the crawler?
The site says it blocked further attempts on Oct. 18, then blocked renewed requests on Oct. 19 and 20. It reports that the requests stopped around 3 p.m. CEST on Oct. 20.
What happens next?
Open3DLab says a fuller log analysis is forthcoming. A Roblox response could address who operated the crawler, why it accessed the sites, and whether the files were retained or used.
Source: open3dlab.com
