AI Deepfakes Become Growing Threat to Global Security and Business

Artificial intelligence has dramatically escalated the age-old tactics of deception and media manipulation in warfare to unprecedented levels. According to recent data from Entrust, new deepfakes are being created at an alarming rate of one every five minutes in 2024. The European Parliament’s projections are equally concerning, estimating that approximately 8 million deepfakes will circulate throughout the European Union this year alone.

These sophisticated technologies have evolved to the point where they can potentially destabilize entire nations without a single shot being fired. The human psychological tendency to respond more quickly to negative news, coupled with our natural inclination to share such information and our poor ability to detect fake content, creates a perfect storm for digital manipulation.

The recent anti-immigrant riots that swept across the United Kingdom provide a sobering example of how rapidly false claims on social media platforms can spiral out of control and transform into real-world violence.

The threats posed by deepfake technology are diverse and far-reaching. Fabricated videos of political leaders making inflammatory statements, doctored audio recordings containing false instructions, and manipulated images can collectively undermine governmental stability and shape public opinion with frightening efficiency.

The corporate world faces similar vulnerabilities. False announcements or fake statements attributed to company boards can trigger significant fluctuations in stock prices and erode investor confidence in minutes. Compounding this problem, several major social media platforms have reduced their content moderation efforts, creating more fertile ground for such deceptive material to achieve viral status.

“Whoever controls information controls the narrative,” goes the saying. While television and newspapers traditionally shaped public perception, today’s social networks reach vastly larger audiences with greater speed. The true danger of deepfakes lies in their ability to pollute the media environment and sow seeds of distrust. When people can no longer trust what they see and hear, their opinions become increasingly susceptible to external manipulation.

Recent years have witnessed an uptick in election interference by authoritarian regimes spreading false information. Russia, in particular, faces frequent accusations of attempting to influence democratic elections worldwide. This trend has sparked widespread concern that deepfakes and voice cloning technologies could fundamentally undermine electoral integrity.

A case in point occurred just before Slovakia’s recent election, when a fake audio clip surfaced allegedly featuring a candidate discussing electoral fraud. Though the recording was later denied, it had already gone viral. Its strategic release during the election’s mandated silence period may have influenced the outcome in favor of another candidate.

Beyond election interference, deepfakes are increasingly becoming weapons in armed conflicts. The Middle East, Ukraine, and the brief war between India and Pakistan have all witnessed numerous deepfakes deployed by all sides involved. Early in Russia’s invasion of Ukraine, a deepfake video showing President Volodymyr Zelensky supposedly demanding his troops surrender circulated widely before being debunked. Security experts suggest this could become the model for future conflicts, where information warfare becomes as critical as physical combat.

The business world faces equal dangers. Strategic companies can be targeted by competitors or adversaries, who can create doubt about products or services through fabricated content. A fake video showing an executive claiming product safety issues can rapidly spread online, damaging customer trust and investor confidence in hours. Stock prices can plummet, partnerships may be questioned, and supply chains disrupted based on entirely fabricated claims.

A recent business survey revealed that eight out of ten executives express concern about how AI-driven false information could harm their operations. The challenge is compounded by the fact that even when fabricated videos are eventually debunked, significant damage often remains. Public perception takes time to correct, and disinformation campaigns frequently continue to insist the original content was authentic and authorities are covering up the truth.

In response to these growing threats, countries worldwide are beginning to implement new legislation. The European Union’s AI Act establishes guidelines for safe and transparent AI use, including provisions addressing AI-generated content risks. Following this approach, Denmark has begun updating its copyright laws to restrict deepfake creation and distribution, giving individuals greater control over their own image and voice representation.

In the United States, the Take It Down Act now requires harmful deepfakes to be removed within 48 hours and imposes federal penalties for distributors.

“Governments and international organizations play a crucial role in combating AI fraud through regulation, policy formulation, enforcement, and international cooperation,” explains Andrius Popovas, Chief Risk Officer at Mano Bank.

Ironically, the most promising solution in combating deepfakes may be AI itself. Advanced artificial intelligence systems can analyze patterns, language nuances, and contextual elements to assist with content moderation, fact-checking, and identifying false information. However, significant limitations remain.

Detecting deepfakes proves substantially more challenging than creating them, primarily because detection systems require extensive datasets of both genuine and fake content, which demand significant human labor to properly label. While public figures like politicians and celebrities may be easier to protect due to abundant reference data, ordinary citizens remain more vulnerable. Furthermore, many detection systems struggle to identify new deepfake generation techniques, often performing well against known methods but failing when confronted with novel approaches.