0

I have an educational app & service that generates quizzes that tests the user on any topic.

I want to expand it so the topic can be any video input.

I have many questions about legality of implementing such a feature:

  1. If the App UI was programmed to have a "create quiz from youtube url" feature (i.e. temporarily download youtube video locally on user device and process it the same way as local video file), would it be illegal to implement such a feature? If it is illegal, what about instead hinting e.g. "you can download video from youtube using this other app, it is your responsibility to check if this is legal for you to do".
  2. The backend service is fully automated and is ignorant to whether or not a user owns the input video. Is it illegal for the service to then process and store the quiz on the backend if it is not shared? i.e. Is it okay to operate a service ignorant to content ownership? (steps can be made to reduce risk of misuse, like user identification, copyright detection algorithm, etc.. but are these things legally mandatory? Because they are prohibitively complex/expensive for me to implement as an individual)
  3. If #2 is fine, would that change if quizzes we're optionally public or shareable for other users to access? My understanding is I would need to have a accessible DMCA takedown contact, if someone wanted to dispute on any of the quizzes.

I will be monetizing this service by selling in-app purchases and displaying ads, and it will operate like a business since it costs money to run the backend.

I intended on making this app for personal use, but have expanded it since I think others might find it useful. It uses AI to curate the important ideas and formulates a quiz. The way I see it, it is no different as if a human was paid to do the same thing. But since it is fully automated, and it seems way too complicated for me to implement comprehensive vetting on input videos, it leaves this "grey" area where I'd be monetizing a service that may be misused without any way of knowing.

The closest thing I could find was questions on "youtube-downloader" apps/websites where it is explained and my interpretation of which is that the service could be taken down and held responsible if it is apparent that is was made primarily as a tool for the user to do such acts and/or the creator/owner of the service knew this. I think this differs from my case for various reasons, e.g. it's not the purpose of the app at all, and there's no reasonable way for me to know if user's are misusing the service, and the video's are all downloaded by the user locally (without touching backend service).

I don't know much about law, but I hope my proposed idea isn't prohibited (in the US or any reasonable legal system). I think it falls closer to a general "fan-appreciation-tool" than "derivative-work/redistributor". What I mean is, it doesn't target any particular work, it is just a general tool. And even if it was "misused" and a user created quizzes on copyrighted work, it is all for his own personal reasons and that probably makes the user even more interested in the original content.

My guess is it would be okay to do this, where it would be similar to how online forums are treated (in terms of sharing quizzes, just need DMCA takedown thing), but with the additional consideration of the "generator tool" which I'm less sure on in particular what was mentioned in #1. Worst case scenario if both those parts are prohibited, I should be able to publish and monetize this app/service without the "sharing quizzes" and/or "quiz from youtube link" features right? But ideally I hope it can still have those features.

1
  • 5
    Sounds like you need advice from a lawyer rather than random strangers on the Internet.
    – bdb484
    Jan 1 at 9:10

1 Answer 1

2

You are going to open several legal 'cans of worms' with this, especially if you monetize it but also if you just publish it ...

  • The legal status of some AI-generated content is currently under dispute. The big generative AI models need training data and that was scraped from the web; they may be systematically violating copyrights.
  • When users are sharing user-generated content, there is not just the copyright issue but also the deletion of illegal content (like child pornography, or revenge porn, or terrorism support). You will need conflict resolution mechanisms.
  • You will have to look at the youtube terms of service when you include them in your app. (You would also be at the mercy of youtube for any API changes, but that's less of a legal problem and more one of business cases.)
  • You may not be able to limit your compliance to the US. Other jurisdictions will expect compliance to their laws, and if you have paying users or advertisers from there they have a lever.
6
  • The legal status of some AI-generated content is currently under dispute. - no it's not under dispute - Ai-generated content is not copyrightable. But the training data might be a copyright infringement, which is a separate issue.
    – Trish
    Jan 1 at 12:47
  • 3
    @Trish, I would answer that less absolutely than you do. There are legislations where databases can be copyrighted even if each individual entry does not rise to copyrightability. Calling a generative AI a 'database' would be a stretch, however. Anyone who wants to base a business case on AI today had better be careful for changes in legal precedent.
    – o.m.
    Jan 1 at 14:17
  • Thanks for the help. From what I can tell, the online part is operated like a forum. It should be fine as long as moderation is taken seriously. It already has "illegal/indecent" content blocking/filters and user moderation system that can report anything that might slip through (but copyright checks is more difficult). My take on that then is that it can safely operate if I include readily available contacts to stop providing content/service, like DMCA takedown.
    – Buretto
    Jan 1 at 18:48
  • The "generation tool" might be a less well understood like you say. In my case, it uses a third party AI model through API. There are now uncountably many apps/services/businesses built off the same AI technology. I'd expect the blame directed towards the model provider in case of any concerns with copyrighted data used in training. For the generated content, the best I can do is offer ways to takedown content if requested or block it from being shared if requested.
    – Buretto
    Jan 1 at 18:49
  • 3
    @Trish: That depends heavily on the kind of AI. For example, LLMs like GPT-x work by stringing together fragments of their training data and have, in extreme cases, found to be reproducing entire paragraphs of text or entire computer programs verbatim. Jan 1 at 21:22

Not the answer you're looking for? Browse other questions tagged .