[notachatbot1234]

> The dataset is released under the Creative Commons BY-SA

How can this be legal? All imagery is taken from (usually) non-free movie trailers.

[mewse-hn]

I had the same question, it seems silly to build a collection of copyrighted content and apply your own copyright to it.

I guess the argument is the same one all the AI people are relying on: I built this collection of fair use material and I am applying my copyright to the product of my work. I wouldn't want to argue that one in court.

[b112]

Yet such court actions are going to happen again and again, until VC backed AI firms either all go bankrupt, or win.

[RIMR]

This dataset is already at risk of existing outside of fair use, but trying to apply your own copyright is pretty much asking to get sued.

[cardiffspaceman]

IANAL: As far as I know, making the compilation earns you a copyright. But for a reader to make a copy, they need licenses from you and from the copyright holders of all the images. So in this case maybe the release notice that you quoted means that there is still all the image copyrights to obtain licenses for.

[dinobones]

IANAL, but https://en.wikipedia.org/wiki/Fair_use is a thing. Not sure if it applies in this case though.

[LogicalRisk]

Fair use is a justification for why copyright restrictions may not apply in a given scenario, not a license to apply new legal restrictions to work you do not own.

[carom]

Datasets and curations are copyrightable. I would think of it as a right to use the curation, not a right to use the actors' likeness.