Lets say I create a program that generates random text based on a seed value.

I then create a secondary program, which uses incremental seeds with the initial program to generate several trillion strings of text. It compares each string against a library of copyrighted works to see if there are matches, and writes down the seed alongside the name of the work if it does.

The end result is a list of seeds that can be used with the original program to generate a specific copyrighted work.

What would then be the legal status of these seeds? Would sharing them be considered copyright infringement?

  • 3
    If you've managed to do that in any meaningful way, before the end of the life of the universe, copyright is going to be the least of your worries...
    – user4210
    May 23, 2019 at 4:19
  • The technical viability isn't really part of the question, but perhaps assume that I have access to Summit for several years to run this.
    – Jivira
    May 23, 2019 at 4:28
  • For some questions, the technical viability simply renders any answer pointless, imho. What you are really asking is "if I have a method of converting a short string into a existing useful copyrighted work, what is the legal considerations of distributing the short string?" and there are better ways of asking that without introducing an impossibility, such as reversing an MD5 hash (thought impossible not that long ago and as such was used for password storage, but today MD5 is woefully broken as a hashing algorithm).
    – user4210
    May 23, 2019 at 4:35
  • Quite simply, the answer would also depend on the jurisdiction - breaking it down, you are asking about a distribution method similar to torrent magnet files (distribute a string which ultimately results in you getting the full work from another means), which are considered a breach of copyright in some jurisdictions and not in others.
    – user4210
    May 23, 2019 at 4:36
  • That's a fair point. The key thing I wanted to avoid was the 'short string' being derived or otherwise being non-independent from the copyrighted work. For example compressing copyrighted text doesn't meaningfully change the legal considerations around it. Neither does creating a magnet link, considering it's just the hash value of a file.
    – Jivira
    May 23, 2019 at 4:49

2 Answers 2


So, you have an infinite monkey machine. More precisely, what you have described is a decompression program (zip/unzip) into which you feed a randomly generated seed.

Since you did not copy the material, you will not have violated copyright law.

However, if the copyright owner were to bring a case they would probably win. That's because cases are decided on the balance of probabilities and the no matter what evidence you provide, their expert evidence based on the infinite monkey theorem, will show it is way more likely that, for a work of any reasonable complexity (like this answer), the number you have was generated by pushing the original work through your algorithm backwards rather than generating it randomly and it happening to correspond with an actual intelligible piece of writing.

You will, of course have copyright in your algorithm.

  • 1
    But he generated the text directly by comparing it to copyrighted material. The copyrighted material was known a priori, and an integral part of the process to select the seed. I can make the same analogy on a photocopier. A photocopier uses an image to charge a plate then randomly sprinkles toner that sticks on the charged areas. The random sprinkling of toner doesn't exempt a photocopy from copyright infringement.
    – user71659
    May 23, 2019 at 5:01
  • The text wasn't generated by comparing it to copyrighted material, as program 1 knows nothing about the texts that program 2 compares its output with. Both the seed and the text are created without consideration for copyright, only for which seeds are saved
    – Jivira
    May 23, 2019 at 5:08
  • @Jivira the fact that you have a comparison step in the flow to specifically reject keys which don't create a useful output complicates the issue, in that all you have really done is taken the long way around to deriving the key from the source material, and that is what the argument against you will be in court.
    – user4210
    May 23, 2019 at 5:24

I believe this is an open issue in copyright law at this time.

First off, I'd like to point out that you might be underestimating the seed. You should expect, on average, a seed which can reproduce a 10kB novel will be roughly 10kB in length. It's basically a fancy encoding of the existing work. You're unlikely to find a 10 digit seed which, when run through your algorithm reproduces a work of Shakespeare. Also, as the original work was used in the production of this seed, it would likely be considered to be a derivative work.

However, because this is a number, the famous controversy over the AACS encryption key is relevant. Hackers identified a key critical to decrypting HD-DVDs and posted it online. The MPAA and AACS LA began issuing cease and desist letters, arguing the number was copyrighted. To my knowledge, no official answer as to whether or not one can copyright a number in this way was ever reached. The issue was handled out of court, with the final result that the AACS went after anyone trying to use the number to break the encryption, but permitting discussion of the number without threat of legal action.

You must log in to answer this question.

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