A man is seen using a mobile phone behind a glass surface. (Jasper James / Getty Images)
An investigation by Consumer Reports has revealed that most of the leading artificial intelligence voice cloning programs lack sufficient barriers to prevent individuals from impersonating others without their consent. Advances in voice cloning AI technology have enabled services to imitate a person’s speech pattern with just a brief sample of audio. Concerns were raised after fake robocalls of a supposed Joe Biden circulated during the Democratic primaries, prompting a subsequent ban on AI-generated robocalls by the Federal Communications Commission.
A recent study of six prominent publicly available AI voice cloning tools found that five had easily circumventable security measures, making it straightforward to replicate a person’s voice without authorization. Detection software for deepfake audio struggles to distinguish between genuine and synthetic voices. Despite being a rapidly evolving technology, the industry lacks significant federal regulations, relying primarily on self-imposed ethical and safety checks.
Voice cloning technology typically involves creating a synthetic audio file based on a person’s voice sample. Many services allow users to upload audio clips, taken from sources like TikTok or YouTube, for replication. The majority of these tools only require a user to confirm consent, with only one service, Descript, implementing a more robust authentication process. While some companies emphasize the potential for misuse and have implemented safeguards, concerns remain about the significant risks associated with unauthorized voice impersonation and deepfake creation.
Although AI voice cloning has beneficial applications, such as aiding individuals with disabilities and facilitating multilingual audio translations, the technology also presents serious threats, including fraud, scams, and disinformation through impersonation of public figures. More research is needed to understand the extent of unauthorized voice cloning and its potential negative impacts on society.
Artificial intelligence is being utilized in audio-related fraudulent activities. For instance, in grandparent scams, perpetrators contact individuals over the phone, pretending that a family member is in dire need of help, such as being kidnapped, arrested, or injured. The Federal Trade Commission cautions that these scams might incorporate AI technology, although they have been occurring before the advent of such technology. Additionally, cloned voices have been employed to produce music without the consent of the original artists. An example of this was a viral song in 2023 that falsely appeared to be by Drake and the Weeknd. Consequently, some musicians have faced challenges in maintaining control over their public image as others release music featuring their voices.
