Google clearly states CSAM is forbidden on their platform.
The user stored CSAM on Google Drive. Regardless whether it is for academic research or not - which in this case, it wasn't. It was for training a model.
Do you really think Google wants to be on the front page of the newspaper saying it allows users to store CSAM?
The user has been caught off guard and that's on them.
And yes, you're not safe either if you store CSAM on Google. Who in their right mind thinks they are?
CSAM is the author's baseless speculation, but I could find no evidence suggesting NudeNet contains any. That it's hosted on GitHub, all its files are on archive.org, and is a widely used data-set 6 years old by now, all indicate that it doesn't contain CSAM.
In general, when a company does something hostile and unreasonable, let's not invent out of thin air facts to excuse their behavior, shall we? Especially since their stonewalling is deliberate.
Google won’t tell me the name of the file—or files—they believe violated their policy. So IT IS baseless speculation. I only know what I was doing prior to the account deletion—and I’m guessing that’s what triggered the issue.
Google should at minimum publish the filenames of the content they consider violations. I have Colab notebooks for everything I’ve downloaded from the internet. With a filename, I could trace the origin and have the image independently reviewed.
I suspect it wasn’t CSAM at all. We know nothing about how their detection system or hashing works. I’d also like to point out that I benchmarked my model against Google’s own commercially available NSFW detector—and mine actually performed better:
https://medium.com/@russoatlarge/benchmark-study-punge-yolov...
As for datasets: COCO is the most widely used dataset in computer vision, with hundreds of thousands of labeled images. One image in the COCO Train2017 set—000000001790.jpg—shows a child brushing his teeth while looking in the mirror in his “birthday suit.” But if Google's system flagged that as CSAM, then every researcher using COCO is at risk.
If Google genuinely cared about addressing CSAM responsibly, they would share the filenames so they could be independently verified and—if needed—removed from circulation. Instead, they’re silent. That silence creates fear, not safety.
You’re making a serious accusation without evidence. I don’t know if the file in question was CSAM—because Google won’t tell me. We just have Google’s word for it.
What I do know is that I used publicly available datasets—like NudeNet and COCO—for benchmarking an NSFW detection model. These datasets have been used by researchers for years. If there was anything questionable in them, that should be addressed openly—not used as a reason to erase someone’s digital life without transparency or due process.
I’m not claiming academic immunity. I’m saying it was unintentional, and I had no idea anything I processed could have been flagged. And if there was something problematic in a dataset, I would absolutely want to identify and report it—but I can’t do that if Google refuses to disclose what was flagged.
This isn’t just about me. When a private company has the unchecked power to destroy someone’s livelihood overnight, it stops being a simple platform—it becomes a de facto utility. There needs to be a path for review, correction, and accountability.
Google clearly states CSAM is forbidden on their platform.
The user stored CSAM on Google Drive. Regardless whether it is for academic research or not - which in this case, it wasn't. It was for training a model.
Do you really think Google wants to be on the front page of the newspaper saying it allows users to store CSAM?
The user has been caught off guard and that's on them.
And yes, you're not safe either if you store CSAM on Google. Who in their right mind thinks they are?
Edit: Grammar