Artificial intelligence (AI) is a branch of computer science that develops systems capable of performing tasks that require human intelligence, using machine-learning algorithms that improve with experience. One of its most relevant applications is voice cloning: a technology that takes audio data, analyzes it, and synthesizes a person’s voice with near-perfect accuracy, recreating timbre, intonation, and vowels in any requested context.
This article analyzes voice, cloning, and artificial intelligence, exploring their regulation and the implications of voice cloning in Mexico. We examine the line between legal and illegal use and how users can protect themselves from fraud.
According to the Royal Spanish Academy, voice is the “sound produced by the vibration of the vocal cords”. In Spain, voice is protected as a privacy right enshrined in the Constitution: recording another person’s voice without consent may constitute a violation of that right.
Applicable legal framework in Mexico
In Mexico, voice is not specifically protected at the same constitutional level as in Spain, but it does find shelter in several statutes relevant to cloning scenarios:
- Federal Copyright Law (LFDA). Protects voice when part of a protected artistic performance or interpretation.
- Federal Personal Data Protection Law (LFPDPPP). Voice, as a biometric datum, qualifies as personal data and, in certain contexts, as sensitive data.
- Federal and local Civil Codes. Personality rights —honor, image, privacy— and civil liability for moral damage.
- Federal Criminal Code. Applicable in cases of fraud, identity theft, or threats through digital means.
Legitimate and illegitimate uses of voice cloning
There are legitimate uses, particularly where the voice holder has given express consent: dubbing, accessibility for persons with disabilities, academic or cultural preservation of voices, and so on. The key is clear authorization and contracts that govern scope of use, duration, economic rights, and revocation.
Illegitimate uses are those without the consent of the voice holder, particularly when fraudulent or defamatory. The most serious cases include:
- Telephone fraud imitating the voices of relatives or executives to request transfers (vishing, variants of the CEO fraud).
- Creation of fake audios attributing statements to public figures (voice deepfakes for disinformation or extortion).
- Use of artists’ voices without authorization to create commercial content.
- Identity impersonation to access financial services using biometric voice authentication.
Intellectual property and AI-generated voice
A relevant debate concerns ownership of works generated with voices cloned via AI. Under current Mexican law, both the LFDA and the LFPPI require the author or inventor to be a natural person. This raises questions about the protection of outputs autonomously generated by AI systems and about the rights of the original voice holder when their voice is replicated without active involvement.
Practical recommendations
- For individuals: verify identity through alternate channels when receiving requests for transfers or sensitive data by phone, even if the voice sounds familiar; agree on key words or questions with family members and coworkers in advance.
- For businesses: reinforce authentication protocols that do not rely exclusively on voice; train staff on CEO-fraud patterns; establish double-validation workflows for payments and transfers.
- For creators and artists: include specific voice-use clauses in collaboration, recording, and advertising contracts; consider registering the voice with INDAUTOR as part of an artistic performance.
- For users of generative voice AI: secure documented consent from voice holders, define purposes and duration, and incorporate mechanisms to revoke and delete samples.
Conclusion
AI-powered voice cloning opens legitimate opportunities in creative, educational, and accessibility areas, but also enables serious risks of fraud, identity impersonation, and affectation of personality rights. In Mexico, the current framework enables action against abuse through the combined application of the LFDA, LFPDPPP, Civil Code, and Criminal Code, although specific regulation addressing voice as a biometric datum and as a personality attribute would be desirable.