Mexico’s New AI Legislation: Regulating Creativity and Data

An important draft legislation has been introduced by Dolores Padierna and other members of Congress. If enacted, it could substantially change the legal landscape of AI in Mexico. The legislation intends to clarify the scope of copyright protection over AI-generated works and also forbid the use of certain AI creations where they may unduly use the image or voice of individuals.

To understand the reasons behind this draft legislation, it is important to mention that the Supreme Court of Mexico recently decided that works created exclusively by artificial intelligence are not subject to copyright protection in Mexico. This decision goes along the same line in other prior decisions of the Supreme Court and which had held that only individuals can be considered as authors in Mexico. In the Court’s view, only human beings can perform the creative activity necessary to create a copyrighted work. Logically, the conclusion is that works where a human being does not intervene are not subject to copyright protection.

Although the Supreme Court has clarified that this ruling applies only to works created without human contributions, a multitude of questions remain. For example, how much human intervention is required in order for a work to be subject to copyright protection? Does a simple editing, cropping, re-sizing suffice or is something more substantial required?

The proposed legislation would amend the Federal Copyright Law to introduce a classification of AI works. Content generated without significant human intervention are classified as “autogenerated works” and exempted from copyright protection. They are in the public domain from the moment of creation. Please do note, however, that a derivative work created from them by a human being could very well be subject to copyright protection.

Hybrid works with a significant degree of human intervention can be subject to copyright protection. To clarify the degree of human intervention required, the draft specifies that the human must actively participate in the process of the creation of the work, making decisions that will substantially impact the result. Human participation must be direct and verifiable.

On this last point, it is foreseeable that rightsholders in these hybrid works will have to maintain logs, documents, and other evidence to show the human participation in the works and that they will likely have to submit such evidence when enforcing the rights over the work.

As with other copyrighted works, hybrid works can be registered with the Mexican Copyright Office. This is optional, as Mexico is part of the Berne Convention, but for these hybrid works registration would be highly advisable.

To obtain the registration, the rightsholder would have to explain to the Mexican Copyright Office the aspects of human intervention of the works and provide proof thereof. The registration would thus provide a legal iuris tantum presumption that the work does have the required level of human intervention necessary to have copyright protection. This would make it more difficult for a potential defendant to argue otherwise in litigation. Equally important, by submitting the evidence to prove the human intervention with the Copyright Office, the rightsholder would be giving it an undisputable date of existence, which would make it harder for a defendant to argue that the evidence of the human intervention in the work was created at a later date.

For this reason, although registration can be advisable for any work, for these hybrid works it will be very important and offer significant advantages when enforcing the copyright against third parties.

In addition to regulating AI works, the proposed legislation intends to make it illegal to use representations of the voice, image, gestures and other identifiable characteristics of an individual without authorization. In addition, the proposed legislation would reform the Federal Criminal Code to use, spread or generate virtual content to supplant or manipulate the identity of a real and identifiable person with the purpose of causing harm or impacting public opinion.

In a recent decision, the Supreme Court had decided that to use the image of an imitator impersonating a famous singer, the permission of the singer would be required. The court had also stated that the right that an individual has to authorize the use of his personal image in Mexico is not only limited to photographs, but could encompass any other depiction of the individual. The proposed legislation seemingly builds on these cases and expressly provides that an AI-generated depiction of an individual would require authorization.

The draft legislation expresses the motive that this is necessary to avoid undue uses of AI-generated contents that may cause harm such as deep fakes. However, there could be unintended consequences upon freedom of speech and seemingly lawful everyday activities. On its terms, a videogame avatar resembling a real person could be potentially unlawful under the proposed legislation. Sharing memes depicting politicians and other public figures could become a source of potential legal liability as well, which could have a very important impact on the use of social media platforms.

When it comes to image rights, one of the most important exceptions is that the image of a person can be used for journalistic purposes without authorization. Here the question is if that same exception is going to apply to ai-generated contents that would otherwise be forbidden under the new legislation. If this is not permitted, the new legislation could have an unintended chilling effect on news outlets and social media platforms.

The draft legislation regarding image rights appears in line with the practices of other jurisdictions and references to legislation implemented recently in Denmark, pointing out that the intended goal is to protect the image of individuals from unauthorized imitations made through AI tools, though it may have unintended consequences. It will be up to the newly-appointed judges of the Federal Judiciary to establish a balance between the right of individuals to be protected from unauthorized uses of their image and the right of the public to be informed and express themselves on matters of public interest.

Lastly, the proposed legislation amends the law regulating the treatment of personal data by private parties. It provides that biometric patterns, algorithmic inferences and behavioral patterns are protected personal data. What this means in practice is that anyone seeking to collect, treat or transfer such data will need to provide a privacy policy. Such policy will have to expressly foresee the purpose of the collection and treatment of the data, as well as the permitted transfers. It also means that any training of AI over this kind of personal data will have to be informed and consented.