Intellectual property and AI-generated content: copyright legal challenges

Intellectual property and AI-generated content: copyright legal challenges

In the digital age, artificial intelligence (AI) has become a powerful tool transforming sectors such as creativity, art, music, and literature, among others. Generative AIs, like ChatGPT or Midjourney, can create original works based on vast amounts of data and complex algorithms, raising several questions about intellectual property and, in particular, copyright law. This is a rapidly evolving topic that still lacks a clear legal framework, creating uncertainty for creators, AI developers, and the companies involved.

Generative AIs are trained using large volumes of data, many of which are protected by copyright. These data may include texts, images, music, and other types of content. However, since AI models process this data to “learn” and generate new works, a debate has emerged on whether using such training data constitutes a copyright infringement.

In many cases, the companies behind these AIs do not disclose which data have been used to train their systems, raising concerns about transparency and the potential violation of content creators’ rights.

Authorship in works created with AI

One of the most complex dilemmas in this field is the question of authorship. Copyright law in many countries, including Spain, states that only natural persons can hold rights over a work. This raises an important question: who is the author of a work created by artificial intelligence? There are different perspectives on this matter:

1. Some argue that the user employing AI to generate content should be considered the author of the work. This argument is based on the idea that, although AI generates the content, the user makes the final decisions regarding the outcome and shapes the creative process.
2. Others believe that the programmer who developed the AI algorithm should hold the rights. This perspective recognizes that artificial intelligence is the product of the programmers’ work, as they design and train the models that enable content creation.
3. Finally, some suggest that the entity commercializing the AI should be the rights holder. These companies control the technology and have a direct interest in the results generated by AI, positioning them as potential copyright holders of AI-created works.

Legal challenges

This lack of clarity regarding the authorship of AI-generated works poses significant legal challenges. Without clear legislation, both AI developers and users of these tools lack the legal protection traditionally provided by copyright law. This creates a legal vacuum in which AI-generated works may not be recognized or adequately protected, potentially falling into the public domain.

Companies developing generative AIs often do not disclose the data used to train their systems. This raises concerns about whether the content used to train AIs has been obtained legally and whether copyright holders should receive compensation. If not, their rights could be adversely affected.

Moreover, there is an ongoing debate on whether authors of original works used in AI training should receive some form of compensation for the use of their creations and be granted certain protections. Otherwise, the rights of these authors may be at risk.

Possible solutions

To address these challenges, various solutions have been proposed to ensure adequate copyright protection. One key solution is to revise copyright laws to specifically address the authorship of AI-generated works. This would include recognizing AI’s capacity to generate content and establishing clear rules on how copyright is assigned in such cases.

Another solution involves developing compensation models for copyright holders whose content is used in AI training. This could include agreements between AI companies and content creators to ensure that authors receive a share of the revenues generated by AI-created works.

Promoting transparency in data usage is also crucial to prevent copyright infringements. Companies should be more transparent about the data they use to train their models and obtain permission from rights holders before using their content for these purposes.