1. Introduction

CVG Law Firm is launching a series focusing on Artificial Intelligence and exploring its legal implications in Turkish law. Rather than engaging in theoretical debates about the legal status or intelligence of AI, our series aims to provide practical insights into assessing the legal challenges arising from the creation and use of AI tools.

The series will be developed in two phases. First, we'll examine the legal status of data processed for the development of AI models and their assessment under Turkish law. This will include a comprehensive review of the relevant data sets in the context of general legislation.

Next, we'll review the regulatory landscape in Turkey, the EU and the US, with a particular focus on recent developments under the EU's Digital Strategy Program. We'll take a look at how these frameworks are shaping AI, particularly in terms of liability law, and unpack the legal implications.

Our first article, "Legal Status of Audio Recordings", begins with an exploration of the use of voice recordings in AI tools, examining their legal infrastructure and protection under civil and intellectual property law. In the second part, we'll look at the protection of Audio recordings in terms of the Personal Data Protection Law and the Turkish Penal Code.

2. Use of Audio Recordings in Artificial Intelligence Tools

Audio recordings serve as a versatile dataset across various sectors, each with distinct applications:

Speech Recognition: AI utilizes voice recordings to convert human speech into text, facilitating systems with voice command features. Examples include virtual assistants, voice-command-operated call centers, and smart home devices.

Natural Language Processing (NLP): This machine learning technology enables AI to interpret, process, and comprehend human language, often used by companies to analyze text and speech data for chatbots.

Speech Synthesis: AI can convert text-based data into human speech, commonly found in software and tools featuring text-to-speech systems.

Emotion Recognition: AI extracts tone, emphasis, and other features from voice recordings to understand emotional states. This is crucial in applications such as customer service and market research for monitoring emotional responses.

Musical Analysis: AI utilizes audio recordings to analyze and recognize pieces of music, aiding in music recommendation systems, automatic instrument tuning, and other music-related applications.

Noise Reduction and Audio Enhancement: By analyzing audio recordings, AI can filter out unwanted noise or enhance audio quality, improving the quality of teleconferences, podcasts, and other audio content.

These technologies find extensive applications across sectors like customer service, marketing, security, education, and entertainment, demonstrating AI's integration with audio recordings. The use of a wide variety of speech recordings is essential to enrich the intelligence of AI in these contexts.

3. Legal Status of Audio Recordings

The human voice is a distinctive personal attribute. It reflects the unique characteristics of an individual and serves as a means of distinguishing him or her from others. Therefore, the human voice is recognised as an inherent aspect of personal identity and is one aspect of the right to privacy. Moreover, the voice is considered to be personal data subject to legal protection.

Depending on the nature of the infringement of this value, different legal guarantees can be applied to protect audio recordings.

3.1 Legal Protections as per Civil Law

After Scarlett Johansson's image and voice were used in "Lisa AI" ad without permission, her attorney filed a lawsuit to take it down. According to Turkish law, individuals have the right to protect themselves against such violations of their right to privacy by filing a claim for protection or a claim for compensation. According to the Turkish Civil Code, the person concerned may pursue the following courses of legal action:

  • Action for a Declaration: This is an action for a declaration that the unauthorised use of a person's voice is unlawful.
  • Preventive Action: This involves seeking to prevent further unauthorised use of your voice. For example, if someone's voice has been recorded without their consent and is to be, or has been, used in advertising, a preventive action can be brought.
  • Action for an Preliminary and Permanent Injunction: This action seeks to stop the ongoing unauthorised use of a person's voice. For example, if a person's voice has been recorded without consent and continues to be used in advertisements, a lawsuit can be filed to stop this infringement.
  • Claims For Recovery Of Earnings And Material And Moral Damages: These are claims for compensation for the harm suffered by individuals. The person may claim the entire amount of the income obtained, as well as the compensation of this damage, provided that he / she proves his / her damage through material and moral compensation lawsuits.

In the context of procedural law, the evidentiary value of audio recordings must also be assessed. The general principle is that evidence must be obtained legally. However, there are exceptions in Turkish law. For example, in a divorce case, a voice recording may be admissible if no other evidence exists or if there are no other means to prove the facts in the recording. Finally, the legality of a voice recording is assessed on a case-by-case basis, considering factors such as the circumstances of the recording and its purpose.

3.2 Audio Recordings and Intellectual Property Rights

In the context of intellectual property, artificial intelligence can infringe the copyright of authors, performers or record companies. The song "Heart On My Sleeve", which was created by cloning of the voices of Drake and The Weeknd by means of artificial intelligence and uploaded to digital music platforms, was removed from the platforms after a complaint by Universal Music Group[1]. It is the subject of another article whether the output of artificial intelligence tools is a work of authorship and, if so, who is the author.

The use of musical works protected by the law, performances (ie. sound recordings) of artists (the owners of the related rights) and fixation of record companies (ie. sound recordings) as input for the commercial use of the AI tool, and therefore its duplication, can lead to the violation of the rights of the author. It is not possible to make a clear judgement yet as to whether the output created by using AI constitutes infringing copyright. If the song created by using AI copyied (by means of verbatim copying or substantial similarity) or imitated in a way that goes beyond inspiration, it can be argued that there is infringement and possibly derivative work. However, if the output, i.e. the musical work, is at the level of inspiration and only the singer's voice is imitated, it may not be easy to prove infringement and obtain copyright protection.

However, there is also a criminal case that can be used by the singer without the need for evidence of infringement. Appropriation of renowned performers is a criminal offence which may be used as a relief to such unauthorised uses. The offence of "reproduction, distribution, dissemination or publication of a work, performance, phonogram or production using the name of a well-known person", listed in Article 71 of the Law on Intellectual and Artistic Works, may be a deterrent to those who make songs using the name of renowned persons through artificial intelligence. The term "using the name(appropriation)" in the law also covers cases in which a connection can be made with the name of a well-known person. The criminal case, which is an extension of the protection of the right of personality under the Civil Code mentioned in the article, makes it possible for the singer/performer to obtain protection by bringing a criminal case against the user of AI product if the artificial intelligence makes appropriate the song for commercial purposes.

3.2.1 Registration of a Sound as a Trademark

It is also possible to register and protect sounds as trademarks, provided that the conditions of Legislative Decree No. 556 are met. Sound trademarks can be divided into (i) sound trademarks consisting of notes, (ii) sound recordings consisting of themes, and (iii) unique sound trademarks that cannot be notated. In order for a sound to be registered as a trademark, it must have the element of distinctiveness and the visual representation of the sound trademark on the portrait expressed in notes and sound spectrograms that cannot be expressed in notes must be clear, precise, detached, easily accessible, understandable, continuous and objective. Legal protection under trademark law will be called into question if an AI tool makes unlawful use of a sound recording or a sound mark registered as a trademark.

3.2.2 Unfair Competition Provisions

A sound may be protected under the unfair competition provisions of Article 84 of Law No. 5846 on intellectual and artistic works, without being considered an intellectual and artistic work or a related right or a trademark. This includes the reproduction for commercial purposes of audio recordings of interviews conducted by a journalist or audio recordings of a podcast. If the AI tool is reproducing these audio recordings, that is, if all or part of the interview recordings are used as input, or if all or part of these recordings are included in the output, the reproduction is suspect and may be the subject of an unfair competition proceeding.

4. CONCLUSION

It is clear from the above that there are many issues that must be taken into account in the use of voice recordings. For this reason, it is necessary to ensure that artificial intelligence databases are used with the above principles in mind. Otherwise, AI manufacturers may be faced with a lot of complaints from voice owners, trademark owners and recording companies. In the US, we've already seen several cases of the illegal use of voice recordings for artificial intelligence training. Therefore, before using a recording, a detailed legal analysis should be carried out and a 'suitability test' carried out, and the recording should be used in accordance with the results of that test. In the light of the information summarised here, it would be appropriate to use audio recordings in AI training after a detailed analysis.

Footnote

1 https://www.bbc.com/news/entertainment-arts-65309313

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.