The entire world has been thrown into a state of dismay at the end of the historic 2026 global electoral cycles. Democracy, and the public sphere, is saturated all around the world with a flood of hyper-realistic, synthetic media produced by AI. Open source generative video tools have developed quickly, outpacing the notion of traditional media literacy defenses. Now audiences are not only being confronted with amateurish fabrication, but also with perfectly-cloned voices, videotaped addresses from world leaders, and entirely-made-up localized crises.
These video generators are open source tools that have proliferated in recent years, and they have now destroyed the public’s faith in what is perceived as the traditional news. In doing so, the world has entered an alarming “post-truth” era in which truth is no longer given. In order to preserve the integrity of digital journalism, tech giants and regulators now have to quickly abandon ineffective policies to moderate content and implement tough cryptographic watermarking on all digital media.
The Death of Visual Verification
The mantra of journalism for more than 100 years has been ‘seeing is believing’. A verified photo or a confirmed broadcast segment had an intrinsic evidentiary value. The advanced synthetic media software has been democratized and this is not the case anymore. There is little or no cost today to generating a highly polished, completely fictional multimedia story. With a sort-of average computer, anyone can create deep fake videos that would take special forensic labs to tell that they are fake.
The consequences for global democracy are catastrophic. In recent campaigns, false video clips mimicking candidates with false statements or fake emergency security announcements appeared on social media platforms in mere minutes at the end of the campaigns. Voters’ trust in these organizations was battered before fact-checkers could study the data and make official withdrawals.
This setting has fostered a dangerous socio-political dynamic in which the truthful and incriminating evidence of corruption can be “forgotten” when the public speaker says that the real media is a malicious deepfake. If all could be simulated, nothing could be believed.
Moving Beyond Algorithmic Detection
For the past several years, technology giants have focused their efforts primarily on developing post-hoc AI detection algorithms. The security tools are meant to detect the subtle visual anomalies, unusual pixel patterns or odd frequencies in the audio data of the uploaded files that indicate that they are from a machine. But this counter-productive strategy has been a failure. There is a cat-and-mouse game between the deepfake producers and AI deep-fake detectors: as soon as a new detector tool becomes available, the producers respond by updating their generative software to evade the parameters of the new tool.
News institutions may find themselves in a defensive position at all times if they have to depend on detection algorithms. Plus, most of these automated verification tools are only available to special users and the ordinary Internet user has to make crucial informational choices on a gut instinct basis.
The limits of the post-hoc detection of structure, demonstrate that the solution to synthetic media is not at the end of the distribution pipeline. Rather, the global media network needs to undergo a paradigm shift towards proactive provenance tracking, in other words, authentic journalism needs to be cryptographically signed as it is created.
Enforcing Cryptographic Watermarking Rules
Hard Cryptographic Watermarking is a strong, mathematical way out of the Post-Truth crisis. Cryptographic watermarks are embedded in the file structure of the media itself and can not be removed with simple editing software, as they would not be visible in the image of the media. This system leverages open, collaborative standards such as those by the Content Authenticity Initiative (CAI) to provide an immutable electronic record for each one of the original media.
The hardware camera uses a secure, encrypted signature that provides a wealth of valuable information; the camera type, the time of the image and the exact location in the world of the image’s geospatial location. Each admin change to the file is recorded on an unchangeable cryptographic chain as it passes between editing desks and publishing systems.
A user can simply browse the content from any website or social network and an easy-to-use browser plug-in or built-in system badge instantly confirms its lineage. In case of unauthorized manipulation of the image and/or production of a synthetic copy, the cryptographic seal automatically invalidates and the consumer is notified of the compromise of the media.
The standard needs to be enforced, which will be a matter of legislative and corporate willpower. Governments should require all generative AI systems to include identifiable synthetic signatures in their product, and the social networks should give preference to and amplify content with verified journalistic signatures. Going forward, expect a lot of opposition from people who want to keep the open-source software free of any restrictions, but it’s the absolute demise of shared truth otherwise. Through the use of cryptographic watermarking, the digital world has the opportunity to create an infrastructure that will enable the verification of authentic journalism, and restore trust.
