When does a program copyright its output?

[BB+]

The title question has arisen in various guises; however, usually "when" is omitted, which can lead to rather sophomoric conclusions (as in common on the Internet). As a philosophical statement, copyright can only apply to things which derive from a human thought process. Generating every possible haiku mechanically does not place them all under copyright. I don't know if it is the best vocabulary, but it often depends on whether something is computer generated, or computer rendered. Some discussion follows.

As a simple example, consider a children's computer program, which, upon a number from 1-9 being pushed, displays a picture and plays some music [all pictures/music are properly licensed]. If one argued that the output of said computer program (or maybe jukebox) was "not subject to copyright" (being output of a computer program), this would be oddish, to say the least. However, here the rendering program doesn't really copyright its output either -- the program acts more as a renderer of copyrighted material than as a generator of anything.

A more complicated example is Google(tm) results. The issue of whether (say) search results themselves are copyrighted is still debated, but I think the answer is no. However, the arrangement of such search results does fall under copyright. But then, is not said arrangement merely a bunch of computer-generated HTML code (output of a program, certainly!), sent to your computer (and displayed by your jukebox browser)? It seems to me that at least parts of this HTML code are the products of the human mind, such as decorative/organisational aspects of the page. These in turn may be dependent on your locale settings, but they do seem to originate from a creative mind in the end. It appears that Google [admittedly in the service agreement, but it seems copyright-coherent] would allow you to state "The current search results for 'foo bar' has foobar.com as #1", but does not allow you to display a screenshot to evince such. Presumably one could build clonedsearchresults.com which would query google.com and arrange the search results in a different manner, and Google could best respond by denying service.

Similarly, if for a popular search Google were to manually modify their algorithmic results, the search results themselves could be copyrightable. A recent case regarding this was with football fixtures, where there was sufficient human involvement to qualify for copyright protection.

Another issue exists with maps, where here Google explicitly notes copyright not only on their presentation of the results of your query, but also on the underlying map data (if applicable, and Google typically leases this from a provider). Again the computer program is essentially acting as a look-up device of said copyrighted map data, and is not generating this itself.

[syzygy]

A more complicated example is Google(tm) results. The issue of whether (say) search results themselves are copyrighted is still debated, but I think the answer is no. However, the arrangement of such search results does fall under copyright. But then, is not said arrangement merely a bunch of computer-generated HTML code (output of a program, certainly!), sent to your computer (and displayed by your jukebox browser)? It seems to me that at least parts of this HTML code are the products of the human mind, such as decorative/organisational aspects of the page. These in turn may be dependent on your locale settings, but they do seem to originate from a creative mind in the end.

Decorations certainly will be copyrighted. Possibly the layout of the page as well, except that the layout of Google doesn't seem very original.

It appears that Google [admittedly in the service agreement, but it seems copyright-coherent] would allow you to state "The current search results for 'foo bar' has foobar.com as #1", but does not allow you to display a screenshot to evince such. Presumably one could build clonedsearchresults.com which would query google.com and arrange the search results in a different manner, and Google could best respond by denying service.

I doubt that Google has a copyright on the ordered search results (and if I understand you correctly you don't think so either).

Similarly, if for a popular search Google were to manually modify their algorithmic results, the search results themselves could be copyrightable. A recent case regarding this was with football fixtures, where there was sufficient human involvement to qualify for copyright protection.

I don't see how fixture lists can be copyrighted, since there is no creative theme underlying the selection of dates and pairings (even though admittedly the barrier for creativity is quite low).

The UK decision has effectively been overruled by C-604/10:

Article 3(1) of Directive 96/9/EC of the European Parliament and of the Council of 11 March 1996 on the legal protection of databases must be interpreted as meaning that a ‘database’ within the meaning of Article 1(2) of that directive is protected by the copyright laid down by that directive provided that the selection or arrangement of the data which it contains amounts to an original expression of the creative freedom of its author, which is a matter for the national court to determine.

As a consequence:

– the intellectual effort and skill of creating that data are not relevant in order to assess the eligibility of that database for protection by that right;

– it is irrelevant, for that purpose, whether or not the selection or arrangement of that data includes the addition of important significance to that data, and

– the significant labour and skill required for setting up that database cannot as such justify such a protection if they do not express any originality in the selection or arrangement of the data which that database contains.

This case was in fact about fixture lists and arose out of the same dispute. From the decision:

19 According to the findings of fact made by the judge at first instance reproduced in the order for reference, the process of preparing the football fixture lists in question in the main proceedings is not purely mechanistic or deterministic; on the contrary, it requires very significant labour and skill in order to satisfy the multitude of competing requirements while respecting the applicable rules as far as possible. The work needed is not mere application of rigid criteria, and is unlike, for instance, the compilation of a telephone directory, in that it requires judgment and skill at each stage, in particular where the computer program finds no solution for a given set of constraints. With regard to the partial computerisation of the process, Mr Thompson states that it does not eliminate the need for judgment and discretion.

(...)

38 As regards the setting up of a database, that criterion of originality is satisfied when, through the selection or arrangement of the data which it contains, its author expresses his creative ability in an original manner by making free and creative choices (see, by analogy, Infopaq International, paragraph 45; Bezpečnostní softwarová asociace, paragraph 50; and Painer, paragraph 89) and thus stamps his ‘personal touch’ (Painer, paragraph 92).

39 By contrast, that criterion is not satisfied when the setting up of the database is dictated by technical considerations, rules or constraints which leave no room for creative freedom (see, by analogy, Bezpečnostní softwarová asociace, paragraphs 48 and 49, and Football Association Premier League and Others, paragraph 98).

40 As is apparent from both Article 3(1) and recital 16 of Directive 96/9, no other criteria than that of originality is to be applied to determine the eligibility of a database for the copyright protection provided for by that directive.

(...)

43 In the present case, it is for the referring court to assess, in the light of the factors set out above, whether the football fixture lists in question in the main proceedings are databases which satisfy the conditions of eligibility for the copyright protection set out in Article 3(1) of Directive 96/9.

44 In that respect, the procedures for creating those lists, as described by the referring court, if they are not supplemented by elements reflecting originality in the selection or arrangement of the data contained in those lists, do not suffice for the database in question to be protected by the copyright provided for in Article 3(1) of Directive 96/9.

Another issue exists with maps, where here Google explicitly notes copyright not only on their presentation of the results of your query, but also on the underlying map data (if applicable, and Google typically leases this from a provider). Again the computer program is essentially acting as a look-up device of said copyrighted map data, and is not generating this itself.

I doubt that photos made by satellites or from airplanes are protected by copyright, and I don't see how the map data can be protected by copyright. At least not in Europe.

However, in Europe databases can be protected by a "sui generis" right. This right protects databases for which there has been "substantial investment" in obtaining or verifying their content. However, this somehow excludes the investment involved in actually creating the data which makes up the content of the database. I find it just impossible to understand .

[User923005]

When humans play a chess game, the output is not copyright, except as a collection. After all, the players are just following the rules of the game.
Chess game annotations, on the other hand can be copyrighted.
"Look and feel" does not have copyright protection (see the legal battles of the spreadsheet wars of the 1980s).

Are we going to come to the point where we say, "Your program says best move is Nxg3 with an eval of +103 centipawns and so does mine. Infringement!"

If I multiply two matrices together is the output exclusively mine? (Hint, optimal multiplication produces a different answer than the straightforward N^3 method).

Computer chess is already smelling poorly because of the existing atmosphere. It seems that an even worse stench is starting to rise.
Perhaps it is not just chess programming, but everything is going that way. I just noticed chess programming because I happened to look there.

[Jeremy Bernstein]

When humans play a chess game, the output is not copyright, except as a collection. After all, the players are just following the rules of the game.
Chess game annotations, on the other hand can be copyrighted.
"Look and feel" does not have copyright protection (see the legal battles of the spreadsheet wars of the 1980s).

Are we going to come to the point where we say, "Your program says best move is Nxg3 with an eval of +103 centipawns and so does mine. Infringement!"

If I multiply two matrices together is the output exclusively mine? (Hint, optimal multiplication produces a different answer than the straightforward N^3 method).

Computer chess is already smelling poorly because of the existing atmosphere. It seems that an even worse stench is starting to rise.
Perhaps it is not just chess programming, but everything is going that way. I just noticed chess programming because I happened to look there.

An excellent demonstration of 'reductio ad absurdum'. Thanks!

Jeremy

[BB+]

I doubt that Google has a copyright on the ordered search results (and if I understand you correctly you don't think so either).

Actually, one point of my post was to try to understand Wolfram|Alpha (not a "search engine", but a computational knowledge engine, to quote the hype). How can they reasonably assert copyright on (some) outputs? I don't claim to know how their engine works, but here is one possibility.

One creates (via human methods, not automated one) for many popular search terms some "metadata" -- this might be a dictionary definition, examples of word usage, related terms, etc. One might expect that a company could easily have 1 million search terms with "metadata" attached, and 10 million is likely feasible. Then, when the user queries X and Y, instead of passing X and Y to the search engine, one passes METADATA(X) and METADATA(Y) to it. My impression is that there is then sufficient human involvement for the result to be argued as copyrightable, in that the inputs were transformed via essentially a lookup into a human-formed table before being passed to the engine proper. But Wolfram is not exactly mainstream in their thinking.

I don't see how fixture lists can be copyrighted, since there is no creative theme underlying the selection of dates and pairings (even though admittedly the barrier for creativity is quite low).

My first understanding was analogous with the example I gave above. The creativity was in a few inputs (e.g. put X vs Y on day Z for it's the Queen's Birthday, or: don't have A at B and C at D on the same day as those matches attract rowdies and it will stress police capacity in the B-D locale -- #15 in your cited opinion says about 200 such club-requests are made every year), and these desiredarata then affected the resulting programmatic output. The ECJ opinion appears to say that such creativity vis-a-vis the data itself [as opposed to selection/arrangement] is not relevant for sui generis databases, as such rights are specifically for things which don't/cannot meet the copyright criterion of originality in the first place (#30 in the ECJ decision notes this in some regard, saying that it is the structure, not the contents, that is protected). Also, although it is seems not to be the main issue, the discussion in #38-39 leads me additionally to think that such input-manipulation should not be regarded as truly "creative" in the case here, being more technical in nature [though as you say, the bar to creativity is traditionally quite low].

Similarly, one can consider a chess program with many user parameters, or perhaps better to have many internal parameters that are (normally) selected automatically in a pre-eval stage. If the author of said program decides to take 1000 didactic positions and override the normal pre-eval computations when one of those was chosen, one would have some basis for a copyright claim on the resulting output from such a position, assuming sufficient human input was apparent in such choices of overriding parameters.

[BB+]

I don't see how the map data can be protected by copyright. At least not in Europe.

Maybe it was unclear what was meant by "map data", by which I think I meant the arrangement of such data as done by the providers to Google. This could embody a variety of things, such as the decision to indicate that business X is located at coordinates Y, or the labelling of boundary lines of parcels. I would suppose that anything that involved human creative input would qualify for copyright.

One US precedent is embodied in the Rockford decision. To quote that:

Rockford Map starts with aerial photographs distributed by the Department of Agriculture. It traces the topographical features from the photographs and draws lines showing townships and sections. Then an employee goes to places where land titles are recorded and reads the books. The employee uses the legal descriptions of the deeds to draw boundary lines indicating the location and size of each parcel. He pencils in the name of the owner. From time to time Rockford Map updates the maps as ownership of land changes.

At least in the US, this arrangement of the data is copyrightable (see #7,13-14) . [As noted in the decision, Rockford placed bogus middle initials in 56 of the parcel owners, of which 54 were copied].

Here Rockford Map made a contribution. Its employees dug through the records and turned the metes and bounds of the legal descriptions into a pictorial presentation. Teasing pictures from the debris left by conveyancers is a substantial change in the form of the information. The result is copyrightable [...] All Rockford Map's copyright covered was the arrangement and presentation of the information, the translation from dusty books of legal jargon to a picture. It was exactly this contribution that Directory Service borrowed.

OTOH, Wikipedia tells me that even the coordinate data itself can be mangled (or "humanified", I guess) if you want to pursue copyright in that direction: [t]he lower bits of the geographic coordinates are mangled in some systematic way so that it would not be obvious to a user of the map, but could be used in a Copyright trial to show that the data had been copied. [...] So don't use digital map coordinates from proprietary sources even if you were to compare every intersection and shape point on the map!

[BB+]

A related football fixture case (Fixtures v Svenska, the former being the outside-the-UK agent of Football Dataco, I think). Quoting #31-36:

31. In the case in the main proceedings, the resources deployed for the purpose of determining, in the course of arranging the football league fixtures, the dates and times of and home and away teams playing in the various matches represent, as Svenska Spel and the Belgian, German and Portuguese Governments submit, an investment in the creation of the fixture list. Such an investment, which relates to the organisation as such of the leagues is linked to the creation of the data contained in the database at issue, in other words those relating to each match in the various leagues. It cannot, therefore, be taken into account under Article 7(1) of the directive.

32. Accordingly, it must be ascertained, leaving aside the investment referred to in the previous paragraph, whether the obtaining, verification or presentation of the contents of a list of football fixtures constitutes a substantial investment in qualitative or quantitative terms.

33. Finding and collecting the data which make up a football fixture list do not require any particular effort on the part of the professional leagues. As Fixtures itself points out in its observations, those activities are indivisibly linked to the creation of those data, in which the leagues participate directly as those responsible for the organisation of football league fixtures. Obtaining the contents of a football fixture list thus does not require any investment independent of that required for the creation of the data contained in that list.

34. The professional football leagues do not need to put any particular effort into monitoring the accuracy of the data on league matches when the list is made up because those leagues are directly involved in the creation of those data. The verification of the accuracy of the contents of fixture lists during the season simply involves, according to the observations made by Fixtures, adapting certain data in those lists to take account of any postponement of a match or fixture date decided on by or in collaboration with the leagues. Such verification cannot, therefore, be regarded as requiring substantial investment.

35. The presentation of a football fixture list, too, is closely linked to the creation as such of the data which make up the list, as is confirmed by the absence of any mention in the order for reference of work or resources specifically invested in such presentation. It cannot therefore be considered to require investment independent of the investment in the creation of its constituent data.

36. It follows that neither the obtaining, nor the verification nor yet the presentation of the contents of a football fixture list attests to substantial investment which could justify protection by the sui generis right provided for by Article 7 of the directive.

I think the Swedish court (and maybe the Finnish too, in the parallel Fixtures/Veikkaus case) found that there was a database right, but as the information was not utilised in the same way by Svenska (see #1-2 in the Application), there was no infringement. [Svenska noted that they got their data from "public domain" sources, such TV listings, rather than copying a list provided by Fixtures (see #4-5 in the Application), but this issue was never really considered by the court, though in the Horseracing/Hill case the ECJ opined that direct access was irrelevant].

Incidentally, it is claimed to cost 11.5 million pounds a year to make the fixtures, while licensing only recoups 7 million of this [the cost to reproduce one team's fixtures is about 300 pounds].

[syzygy]

I doubt that Google has a copyright on the ordered search results (and if I understand you correctly you don't think so either).

Actually, one point of my post was to try to understand Wolfram|Alpha (not a "search engine", but a computational knowledge engine, to quote the hype). How can they reasonably assert copyright on (some) outputs? I don't claim to know how their engine works, but here is one possibility.

One creates (via human methods, not automated one) for many popular search terms some "metadata" -- this might be a dictionary definition, examples of word usage, related terms, etc. One might expect that a company could easily have 1 million search terms with "metadata" attached, and 10 million is likely feasible. Then, when the user queries X and Y, instead of passing X and Y to the search engine, one passes METADATA(X) and METADATA(Y) to it. My impression is that there is then sufficient human involvement for the result to be argued as copyrightable, in that the inputs were transformed via essentially a lookup into a human-formed table before being passed to the engine proper. But Wolfram is not exactly mainstream in their thinking.

I think to determine whether a particular person or legal entity can claim copyright on a particular output of a machine, you have to consider whether copyrightable expression in that output is attributable to that person or entity in the sense that that person or entity "created" the expression.

If you take Microsoft Word, then it is clear that its textual output is not attributable to Microsoft, but to the person hitting the keys that led to the textual output (unless that person was merely copying a text, obviously). On the other hand, Microsoft theoretically might be able to claim copyright on printed output if it uses a font that Microsoft has the copyright on. That copyright would obviously not protect against copying of the textual content, but would protect against copying of the shapes of the letters. (However, I suppose that Microsoft has implicitly (or even explicitly) given a license to copy the shapes of letters, as long as you don't copy the whole font for use as a font.)

In the case of Google search results, Google obviously has the copyright on its logo and maybe some other graphical aspects. However, the search results themselves and their ordering are the result of the application of a set of uncopyrightable rules (applied on what Google gathered by crawling the web and a search query input by the user). What Google contributed to this is therefore not copyrighted, independent of whether Google crawled the web and produced the search results by applying these rules in a fully automated manner, or whether (employees of) Google did everything manually. At least this is how I understand European copyright law. It seems that copyright in common law countries traditionally takes into account the "blood, sweat and tears" that went into producing a work. These factors are irrelevant for the continental version of copyright (which view has been adopted by the EU). More or less in return, the EU has created this "sui generis protection" for databases.

If metadata produced by (employees) of Wolfram is copyrightable, and if their search output shares copyrightable expression with that metadata, then I guess Wolfram can claim copyright. I'm not very familiar with their engine, but I kind of doubt that it would work (in Europe). Certainly if the metadata consists merely of facts and functional data, there is nothing copyrightable there (whether input by a human or not).

The ECJ opinion appears to say that such creativity vis-a-vis the data itself [as opposed to selection/arrangement] is not relevant for sui generis databases, as such rights are specifically for things which don't/cannot meet the copyright criterion of originality in the first place (#30 in the ECJ decision notes this in some regard, saying that it is the structure, not the contents, that is protected).

Yes, but note that #30 is not about the "sui generis protection" for databases. I guess what #29 and #30 express is that "database copyright" only protects the selection or arrangement of the elements of the database, not the elements themselvse. This does not mean that the elements themselves could not be protected by copyright. Computer programs stored in a file system are elements of a database, but protected by copyright for being a computer program, not for being an element of a database.

Also, although it is seems not to be the main issue, the discussion in #38-39 leads me additionally to think that such input-manipulation should not be regarded as truly "creative" in the case here, being more technical in nature [though as you say, the bar to creativity is traditionally quite low].

Yes. Of course there can also be "technical creativity", but such creativity is not "creative" in the sense of copyright. The shape of a chair can in principle be protected by copyright, but not if this particular shape was designed to allow stacking of chairs. In the latter case, the shape is not "creative" in the sense of copyright. This should not be misunderstood as a judgement of the work of the designer. His design might be patentable. Chess players playing to win might play games full of creativity, but such creativity is not "creative" in the sense of copyright.

Similarly, one can consider a chess program with many user parameters, or perhaps better to have many internal parameters that are (normally) selected automatically in a pre-eval stage. If the author of said program decides to take 1000 didactic positions and override the normal pre-eval computations when one of those was chosen, one would have some basis for a copyright claim on the resulting output from such a position, assuming sufficient human input was apparent in such choices of overriding parameters.

Those 1000 didactic positions might very well represent a creative selection of positions. If you add them to a chess engine, then the resulting source code and object code will be protected by the copyright on this selection.

I doubt however that the copyright on this selection somehow extends to the evaluation function as some kind of independent work. I would say there needs to be some form of sensory observability before you can speak of "expression", and this seems to be absent in case of the evaluation or the functionality of the chess engine. What one observes is that it plays the game of chess, but the game of chess is not copyrighted.

I also doubt that the copyright on the selection somehow extends to the output of the engine. The numbers and moves output by the engine are just numbers and moves. The engine will not reproduce the selection of 1000 didactic positions in its output (unless you make a special function for outputting exactly that selection of positions).

[syzygy]

I don't see how the map data can be protected by copyright. At least not in Europe.

Maybe it was unclear what was meant by "map data", by which I think I meant the arrangement of such data as done by the providers to Google. This could embody a variety of things, such as the decision to indicate that business X is located at coordinates Y, or the labelling of boundary lines of parcels. I would suppose that anything that involved human creative input would qualify for copyright.

But this seems to be merely a bunch of facts, represented as accurately as possible, and selected for being as complete as possible within technical and/or budgetary constraints.

I think in Europe Rockford Map would have to hope for the sui generis right to apply.

The mangling of coordinates is useful to have evidence of copying. It will not help to turn the data into something copyrightable. In the US, this data seems to be copyrighted, mangled or not (but you still need to be able to show copying to succeed in court). In Europe, I don't think it is copyrighted (so it won't help to prove that it was copied, at least not for copyright). Again, the sui generis right might help.