The Rising Threat of Synthetic Media: Insights from VerifyLabs.AI
Synthetic media, particularly in the form of deepfakes, has rapidly transitioned from mere technological curiosity to a growing threat within just a few years. The ease of producing convincing deepfakes has made them a mainstream concern, with simple apps and readily available audio clips enabling malicious actors to create realistic forgeries in a matter of seconds. This development raises significant alarm bells, as the ramifications extend far beyond political misinformation. Serious impacts include corporate fraud escalating into tens of millions of dollars, breaches of biometric security systems, and a troubling surge in non-consensual intimate imagery, a phenomenon largely underreported. As regulatory frameworks in regions such as the EU, UK, and US strive to address these challenges, the pressing question facing businesses and digital platforms has shifted away from whether synthetic media poses a risk, to how quickly they can implement effective measures to authenticate what they encounter online.
In discussions with Ruth Azar-Knupffer, co-founder of VerifyLabs.AI, various aspects of this evolving landscape were explored. Azar-Knupffer elucidated the ongoing “detection arms race,” the emerging regulatory environment, and her conviction that treating verification as infrastructure rather than a product solely owned by one enterprise is crucial in staying ahead of the threat posed by synthetic media.
A Proliferation of Synthetic Media
The prevalence of synthetic media, particularly on social media platforms, has dramatically shifted in recent years. Azar-Knupffer noted that estimates suggest around 500,000 deepfakes were shared across various platforms in 2023, with projections anticipating this number to rise to roughly 8 million by the end of 2025—representing a staggering growth rate of approximately 900% annually. The rise of voice deepfakes is especially notable, with Pindrop reporting an increase of nearly 700% in the year-on-year occurrence of voice manipulation.
While the figures are undeniably striking, Azar-Knupffer emphasized that the transformation in recent years is less about sheer volume and more about accessibility and quality. Three years ago, creating a convincing deepfake required considerable technical skill and resources. Now, the process can be executed with just a free app and a short audio clip scraped from various sources. This stark evolution has made it nearly impossible for the average viewer to differentiate between high-quality fakes and genuine content. Studies have consistently shown that most people can no longer reliably spot a deepfake, marking a significant shift from being able to “spot the fake” to adopting an “assume nothing” mindset.
Keeping Pace with Detection Technology
As the sophistication of deepfake generation continues to evolve, the capacity of verification technologies to keep pace is inherently complex. Azar-Knupffer underscored that while combating deepfakes is viable, it does not conform to a simple “solved” paradigm. Instead, it resembles an ongoing arms race akin to anti-virus software or spam filtering. The objective is not to definitively eradicate deepfakes, but to maintain an operational edge through continuous innovation.
A crucial error that organizations make is concentrating on visible artifacts associated with deepfakes, such as irregular blinking or distorted facial features, which tend to diminish with each new technological advancement. Instead, a layered approach is imperative, employing detection models that interpret signals beyond human perception and integrating provenance—information regarding the original content’s creation and any subsequent alterations. Regulations like C2PA and emerging content-labeling standards are essential as verification practices need to occur upstream, at the content’s source.
Consequences Beyond Political Misinformation
The implications of synthetic media extend far beyond politics. Azar-Knupffer articulated that while political narratives often dominate headlines, the real harm manifests elsewhere, primarily through various forms of fraud. For instance, the infamous Arup case in which a finance employee in Hong Kong was induced to transfer approximately $25.6 million following a deepfake video call highlights the looming danger posed by synthetic media in corporate settings. Furthermore, projections indicate that generative AI-enabled fraud in the US could escalate from approximately $12 billion in 2023 to an alarming $40 billion by 2027, according to Deloitte.
Additionally, the threat to personal identity is significant. Deepfakes have been effectively used to circumvent biometric security measures, with reported bypass attempts surging more than 700%. The rising prevalence of non-consensual intimate imagery, especially targeting women and girls, adds a further layer of complexity and urgency to the conversation surrounding synthetic media.
Collaborative Responsibility in Detection
A critical question that arises is the distribution of accountability in the detection of synthetic media: should platforms, users, or independent verifiers hold this responsibility? Azar-Knupffer argued that no single entity can be solely accountable; otherwise, the issue may fall through the cracks. Instead, platforms must implement detection and provenance strategies due to their extensive reach in disseminating content. Additionally, independent verifiers play a vital role, ensuring assessments are unbiased and not conflicted by content ownership. Users, too, need accessible tools accompanied by a foundational literacy that promotes active questioning.
She emphasized that verification should be conceptualized as infrastructure, analogous to road safety, where manufacturers, regulators, and users all share responsibility. Failure to acknowledge this shared responsibility leaves gaps in protections against synthetic media.
Erosion of Trust
The advent of synthetic media technology poses significant risks to trust in authentic content, accentuating concerns that resonate deeply within modern society. Azar-Knupffer expressed that the acknowledgment of the existence of convincing fakes alters the fundamental ground of trust. The challenges are not confined to the propagation of misinformation; the “liar’s dividend”—where genuine occurrences are dismissed as deepfakes—represents a particularly corrosive consequence, eroding public trust in all forms of media.
Looking Ahead: Preparing for the Future
The landscape of synthetic media is continuously changing, and organizations must proactively prepare for imminent threats. Azar-Knupffer predicted the emergence of real-time, interactive deepfakes capable of simulating live conversations, further complicating detection and verification processes. Additionally, fully synthetic identities poised to bypass existing Know Your Customer (KYC) and biometric checks signal a pressing need for vigilance.
Organizations are encouraged to cultivate a mindset of operational risk management regarding synthetic media rather than treating it merely as an awareness issue. This involves incorporating verification and provenance into essential business workflows and adopting measures such as out-of-band authentication for critical transactions.
In summary, the responsibility to combat the challenge of synthetic media cannot be relegated to one party alone. It requires a collaborative approach involving businesses, technological platforms, and everyday users. The potential dangers are real, and organizations that embrace proactive risk management, invest in verification technologies, and empower users with the skills to question authenticity will be best equipped to navigate the murky waters of synthetic media.