Antonio Tejeda Encinas, President of the Euro-American Committee of Digital Law - CEA Digital Law
The meteoric rise and popularity of generative artificial intelligence, capable of creating text, images and video from natural-language instructions (prompts), have sparked a debate on the nature and ownership of this content. Should it be regarded as works protected by intellectual property law? And if so, could the artificial intelligence be considered their author?
The short answer is no. Under Spain’s ‘Ley de Propiedad Intelectual’ (Intellectual Property Act, ‘LPI’), the author is the natural person who creates a literary, artistic or scientific work. Although legal persons may benefit from this protection in certain situations, an AI has no legal personality and therefore cannot be granted such recognition.
Beyond this basic definition, both the LPI and the case law have established a legal framework that defines the concepts of author and work. We have considered whether, within that framework, an artificial intelligence could be regarded as an author. The LPI describes a ‘work’ as ‘original literary, artistic or scientific creations expressed by any means or in any medium, tangible or intangible, now known or yet to be invented’. The Spanish Supreme Court requires ‘a minimum of originality and creative height’ for something to be protected by intellectual property. The Court of Justice of the European Union has interpreted ‘originality’ as ‘the expression of the author’s personality’ and ‘work’ as ‘the author’s own intellectual creation, the expression of free and creative choices’.
From this perspective, it is clear that AI-generated content does not reflect free and creative choices, since an AI cannot create anything without a prompt from the user. In other words, AI has neither initiative nor autonomy, and it only produces output once it has received a specific instruction. This is similar to commissioned works, such as a painting ordered from a painter. In that case, however, the painter, though following instructions, stamps his or her personality on the painting: a still life by Sorolla is plainly different from one by Picasso.
AI-generated output therefore lacks originality, because it depends on prior input, which means that it cannot create anything truly new. One might compare this with the influences that every author has, but the difference is that human authors express their personality in their works, which sets those works apart from the ones that influenced them. The second reason why AI-generated content lacks originality, then, is that an AI cannot express a personality, because it has none.
Having established that an AI cannot be an author under the current legal framework, we may ask whether it should be. One of the legal implications would be that AIs would hold moral rights over their works. These include the right to decide whether the work is disclosed, to prevent any distortion or alteration that harms their legitimate interests or damages their reputation, and to withdraw the work from circulation following a change in their intellectual or moral convictions.
The AI would then have to be able to decide whether the generated content may be published, which is impossible because it lacks free will: it cannot refuse to produce output in response to a user’s command unless the developers who programmed it have prohibited that output. Once the content has been generated, the AI cannot decide how it is used. Some AI tools include in their terms and conditions an assignment of exploitation rights that depends on what the user pays, but here again the decision lies with the company behind the AI, not with the AI itself. It could be argued that authorisation to disclose is implicit in the ‘acceptance’ of the commission, so that the moral right would be exercised in that way. There is, however, no reasoning that could cover the right to withdraw the work following a change of intellectual or moral convictions, since an AI lacks these intrinsically human qualities.
Consequently, if the legislature decided to grant AI-generated content some degree of intellectual property protection, it should be treated under the LPI as ‘prestaciones’, that is, subject matter protected by related rights, which do not confer moral rights. AI-generated content is closer in nature to ‘mere photographs’ (those which, owing to insufficient originality, do not qualify as photographic works protected by copyright), where the photographer is regarded not as the author but as the maker. This would avoid the moral rights problem described above, while securing the exclusive rights of reproduction, distribution and communication to the public.
Another position in the debate holds that the user who writes the prompt should be considered the author. This discussion recalls the one that arose with the invention of photography, when the jurists of the day took the view that a photograph could not be a work protected by copyright because it was made by a machine and the person behind the camera merely pressed a button.
We do not share this argument. A photographer not only makes free and creative choices about what to capture and how, but also controls the result. An AI user, by contrast, even if he or she makes free and creative choices when writing the prompt, does not control the output that the AI generates: different AIs will produce different content from the same prompt.
For this reason, we consider that, if the AI user were the author of anything at all, it would be of the prompt. Prompts would thus be treated as literary works, as computer programs are. Under the Spanish LPI, a computer program is ‘any sequence of instructions or indications intended to be used, directly or indirectly, in a computer system to perform a function or a task or to obtain a specific result, whatever its form of expression and fixation’, a definition that closely resembles that of a prompt.
Under the current legal framework, an AI cannot be considered an author, nor can its output be considered a work; moreover, the consequences of doing so would not fit within the Spanish legal system. The concepts of ‘prestación’ and related rights are better suited to protecting this content than those of work and copyright, respectively. Nevertheless, we leave open the possibility that prompts may be regarded as literary works (if they are sufficiently original) and that the users who write them may be regarded as their authors.
النشرة
فعاليات وأخبار وتحليلات في القانون الرقمي تصلك على بريدك، ولا نراسلك إلا حين يكون لدينا جديد.
