What if a loved one called you in distress, begging for money, only for you to realise later that it wasn’t them at all, but an AI-generated voice? Cybercriminals are now using AI-powered voice cloning to impersonate individuals and scam unsuspecting victims. ALSO READ: How AI is reshaping customer service in Rwanda How AI voice cloning works AI voice cloning is a form of deepfake technology that enables artificial intelligence to mimic a person’s speech patterns, accent, and tone using short audio samples. Daniel Etukudo, a Computer Vision Engineer and AI programmer, explains that AI voice cloning relies on deep learning models such as text-to-speech (TTS) and generative AI to replicate human speech. “The process involves four key steps: data collection (gathering voice samples), feature extraction (analysing pitch, tone, and speech patterns), model training (teaching AI to mimic the voice), and synthesis (generating speech based on new text input),” he said. Some of the most advanced AI models for voice cloning include OpenAI’s Whisper, ElevenLabs, and Meta’s Voicebox. “A few years ago, cloning a voice required hours of recorded speech and expert tuning. Now, AI can replicate a voice using just a few seconds of audio,” Etukudo added. ALSO READ: How Rwanda is regulating artificial intelligence AI voice cloning in Rwanda AI voice cloning is being used in Rwanda across various industries, including education, media, and customer service. “In education, AI-generated voices help translate materials into local languages like Kinyarwanda. In media and entertainment, Rwandan news agencies and filmmakers can use AI voice synthesis to create high-quality content. Customer service centers also deploy AI-generated voices to enhance support services,” said Etukudo. However, AI voice cloning also raises serious security concerns. “One of the biggest threats is voice phishing, or ‘vishing,’ where scammers clone the voice of a trusted figure—such as a CEO, government official, or family member—to trick victims into transferring money or revealing sensitive information,” he explained. Another concern is the use of cloned voices for fake emergency calls. “Scammers can impersonate kidnapped relatives to demand ransom or misuse AI-generated voices of public figures to spread misinformation,” he added. Remy Muhire, a software engineer and founder of Pindo app, emphasised that AI voice cloning is already being used in Rwanda. ALSO READ: Where does Rwanda stand in integrating AI, technology? “At Pindo, we help organisations maintain brand identity by creating custom AI voices. For example, a bank can clone the voice of its most trusted call center representative to ensure consistent customer service,” Muhire explained. However, he acknowledged the risks: “To avoid scams, people should avoid sharing voice recordings publicly, be cautious of unknown calls, and use security codes instead of voice authentication.” Despite the risks, Muhire noted the benefits of AI voice cloning in accessibility and entertainment. “For example, Stephen Hawking used a computerised voice to communicate. Today, AI can recreate a person’s natural voice, allowing individuals with speech loss to speak in their own voice rather than a robotic one,” he explained. AI voice cloning also enhances industries such as filmmaking, audiobooks, and customer service by providing consistent and engaging voices, Muhire said. How to protect yourself from AI voice cloning scams Muhire stressed the importance of proactive measures against AI voice cloning scams. He advised people to limit voice exposure by avoiding public sharing of voice recordings, being cautious of unknown calls, and opting for security codes instead of voice authentication. This includes being mindful of sharing voice notes on social media or other platforms where they can be misused. Using AI detection tools is another step. Platforms such as Resemble AI and Deepfake Detection AI can help identify cloned voices before they cause harm. While AI voice cloning offers many benefits, Muhire warned that its misuse for fraud continues to rise. He also stressed the importance of enabling multi-factor authentication (MFA). “Instead of relying on voice-based authentication, individuals should use passphrases and other security measures that AI-generated voices cannot bypass,” he said. He added, “Another critical precaution is verifying unexpected voice requests. If someone claiming to be a friend, relative, or boss urgently requests money or sensitive information, it is essential to confirm their identity through a video call or by asking a personal question only they would know. “Scammers can manipulate AI voice technology to impersonate trusted individuals, making verification a crucial step in preventing fraud. Additionally, raising awareness is key to combating AI-powered scams. “Educating the public about deepfake scams helps people recognise and avoid potential threats. As AI voice cloning advances, public awareness and cybersecurity best practices will be vital in ensuring people stay protected, Muhire added.
