The Rise of Deepfake Audio: A New Threat to Democracy?

A recent video of US President Joe Biden speaking to Ukrainian President Volodymyr Zelensky has raised concerns among experts and the public about the potential misuse of deepfake audio technology. The video, which was posted on YouTube by a channel called "Ukraine News", shows Biden congratulating Zelensky on his re-election and promising to support Ukraine's sovereignty and territorial integrity. However, the audio of Biden's voice was not authentic, but rather a synthetic imitation created by a deepfake algorithm.

Deepfake audio is a form of artificial intelligence that can generate realistic-sounding speech from text or voice samples. It can also manipulate existing audio recordings to change the content or the speaker's identity. Deepfake audio can be used for various purposes, such as entertainment, education, or accessibility. However, it can also pose serious risks to democracy, security, and human rights, especially when combined with deepfake video or images.

Some of the potential harms of deepfake audio include:

• Spreading misinformation and propaganda: Deepfake audio can be used to create fake news or political messages that can influence public opinion or elections. For example, in 2019, a deepfake audio clip of former US House Speaker Nancy Pelosi was circulated online, making her sound drunk and incoherent. Similarly, in 2018, a deepfake audio clip of former US President Barack Obama was created by comedian Jordan Peele, warning about the dangers of deepfakes.
• Impersonating or blackmailing individuals: Deepfake audio can be used to impersonate or blackmail celebrities, politicians, business leaders, or ordinary people. For example, in 2019, a UK-based energy company was scammed out of $243,000 by a fraudster who used deepfake audio to mimic the voice of the company's CEO. Similarly, in 2020, a Canadian woman was harassed by a stalker who used deepfake audio to mimic the voice of her ex-boyfriend.
• Undermining trust and credibility: Deepfake audio can be used to undermine the trust and credibility of individuals or institutions. For example, in 2020, a deepfake audio clip of Indian Prime Minister Narendra Modi was used to spread false information about the COVID-19 pandemic. Similarly, in 2021, a deepfake audio clip of World Health Organization Director-General Tedros Adhanom Ghebreyesus was used to spread false information about the COVID-19 vaccine.

How can we detect and prevent deepfake audio?

The detection and prevention of deepfake audio is a challenging task that requires collaboration among researchers, policymakers, media outlets, and civil society. Some of the possible solutions include:

• Developing and deploying technical tools: Technical tools such as machine learning algorithms, digital watermarking, or blockchain can help detect or verify the authenticity of audio recordings. For example, Microsoft has developed a tool called Video Authenticator that can analyze videos and provide a confidence score on whether they are manipulated or not. Similarly, D-ID has developed a tool called Live Voice Authentication that can verify the identity of a speaker in real time using biometric data.
• Establishing and enforcing legal and ethical frameworks: Legal and ethical frameworks such as laws, regulations, codes of conduct, or standards can help deter or punish the malicious use of deepfake audio. For example, in 2020, California passed a law that bans the use of deepfake audio or video in political campaigns without disclosure. Similarly, in 2021, the European Union proposed a regulation that requires online platforms to label synthetic or manipulated content and remove illegal or harmful content.
• Raising awareness and educating the public: Awareness and education campaigns can help inform and empower the public about the risks and opportunities of deepfake audio. For example, in 2020, the Partnership on AI launched a project called Reality Defender that aims to educate the public about how to spot and report deepfakes. Similarly, in 2021, UNESCO launched a project called Truth Never Dies that aims to promote media literacy and critical thinking among young people.

Deepfake audio is a powerful and emerging technology that can have positive or negative impacts on society. It is important to be aware of its potential uses and abuses and to take action to ensure its responsible and ethical development and deployment.