The Federal Trade Commission (FTC) is seeking innovative solutions to address the growing threat of AI voice cloning, also known as audio deepfakes. To incentivize participation, they are offering a $25,000 reward to the best solution. The proliferation of online services that provide easy-to-use voice cloning tools has raised concerns about potential abuse in cyberattacks.
One of the major concerns is the impersonation of CEOs to deceive finance departments into transferring money to attacker’s accounts. This kind of fraud could also trick friends and family into sending money to individuals posing as their loved ones. Additionally, the livelihoods of performing artists are at risk if this technology continues to advance.
The FTC is calling for submissions until January 12 that focus on prevention, monitoring, and evaluation of AI-based voice cloning fraud. The goal is to shield consumers, creative professionals, and small businesses from harm. If viable ideas do not emerge, it may serve as a warning to policymakers about the necessity for stricter limits on the use of this technology.
Submissions will be evaluated based on feasibility, resilience to technological changes, and considerations of liability and responsibility. The top prize is $25,000, while one runner-up will receive $4,000, and three honorable mentions will be rewarded with $2,000 each. Furthermore, organizations with 10 or more members will receive a cashless recognition award.
AI voice cloning has already been used effectively in various cases. For instance, a UK energy company lost $243,000 when its CEO was tricked into sending money to a fraudulent supplier. Banks have also been targeted, and romance scams have become prevalent, with criminals mimicking celebrities and loved ones to dupe victims into giving them money.
To perform voice cloning, an AI model requires sufficient training data to understand the nuances of an individual’s voice, including tone, pacing, and inflection. Celebrities and public figures are particularly vulnerable given the abundance of voice recordings available online. However, even non-celebrities and children have enough online material to train a model effectively.
While there are freely available voice cloning tools, paid solutions like Microsoft’s VALL-E model generate more convincing clones. These technologies are still in the early stages of development and are likely to improve in accuracy and effectiveness over time. The FTC’s reward initiative aims to combat this growing threat and protect individuals and businesses from the harm caused by AI voice cloning.