Roblox downloaded 145GB models from us (including adult content) without asking
Roblox is reported to have bulk-downloaded 145 GB of user-uploaded 3D models — including adult content and non‑commercially licensed assets — from the open3dlab site via an AWS-based crawler that ignored robots.txt and the site’s terms of service. Commenters debate whether large-scale scraping of publicly accessible content is ethically acceptable or should be seen as copyright infringement, especially when used to train commercial AI or moderation models. The exchange broadens into concerns about power imbalances between small creators and big platforms, the practical limits of technical defenses against crawlers, and the erosion of earlier “good manners” norms on the web.
Overall reaction to Roblox scraping
- Many see large-scale scraping of community assets without consent as disrespectful, especially when robots.txt, site ToS, and non‑commercial licenses are ignored.
- Others argue that anything publicly downloadable on the web will be scraped, and site operators should assume that reality and engineer accordingly.
- Some say the real “news” is not the bytes transferred but a big commercial platform quietly harvesting fan content in violation of stated terms.
Licensing, copyright, and “theft”
- One side: downloading non‑commercial / no‑derivatives content for commercial AI training is a clear license violation and effectively “industrial-scale theft,” even if law is lagging.
- Opposing side: merely downloading from a public site is typically legal; copyright only bites when you redistribute or commercially exploit the result, and courts may deem training “fair use.”
- Several note that big companies can and do enforce ToS in court, while small sites rarely can. There’s disagreement on whether terms or robots.txt have legal force versus being mere norms.
Adult content and Roblox’s kid audience
- The article’s emphasis on adult models triggers debate.
- Some see it as emotional framing; the same dataset could plausibly be used to train NSFW detectors rather than generate porn.
- Others say, given Roblox’s child‑heavy audience and prior controversies, even incidental scraping of adult content looks reckless and undermines trust.
Alleged hypocrisy of Open3DLab
- Multiple commenters point out that Open3DLab hosts many ripped game assets and even promotes tools for extracting them.
- The site itself acknowledges operating in copyright “grey areas” and complying with takedowns.
- Critics say complaining about corporate scraping while distributing infringing models is hypocritical; defenders argue that human‑scale fan sharing is not morally equivalent to corporate, industrial‑scale ingestion.
Defenses against scraping
- Suggestions include blocking or tarpitting cloud provider IP ranges, but others note this breaks office VPNs, legitimate crawlers, and doesn’t stop residential proxies.
- Robots.txt is described as “politeness,” not protection; real controls would be rate limits, CAPTCHAs/turnstiles, paywalls, logins, or selective indexing.
- Some argue that small sites can’t afford advanced bot‑mitigation and that constant bot load has turned the web into a tragedy of the commons.