in OpinionMEMBER EXCLUSIVE
Sepia-toned, disordered, and indistinct visuals of the Allied forces disembarking on the shores of Normandy in 1944 are stirring and important partly because we understand their authenticity. Images fabricated by artificial intelligence undermine this collective grasp of reality.(Image credit: Universal History Archive via Getty Images)Share this article 0Join the conversationFollow usAdd us as a preferred source on GoogleSubscribe to our newsletter
Generative artificial intelligence (AI) is blurring the boundary between what is real and what is imagined to such an extent that visual perception is no longer a reliable indicator of truth. We require a societal and legal structure that differentiates authentic visual records from those produced by AI, alongside technological advancements, such as universal “AI watermarks,” to enable viewers to instantly discern genuine imagery from fabricated content. Without such an infrastructure, we risk losing the confidence that real-world photography inspires, which would be detrimental to democracy.
On June 6, 1944, Allied forces launched an assault on the beaches of Normandy. The photographic evidence that emerged — grainy, unfocused, disorganized — did more than merely chronicle events; it actively shaped them. For millions who would never witness the conflict firsthand, these images became synonymous with the war itself, serving as tangible proof of sacrifice, bravery, and a unified objective. They transcended linguistic barriers, bridging the gap between the observer and the actual occurrence.
The same can be said for other pivotal moments in history. The solitary individual confronting tanks in Tiananmen Square. The falling man during the World Trade Center attacks. The deceased form of 3-year-old Alan Kurdi washed ashore on a Turkish beach. These visual records are not simply historical documents; they are cultural landmarks. They constitute a shared visual foundation upon which public comprehension — and frequently, political will — is constructed. They empower communities to align emotions, judgments, and actions on a broad scale.
However, what transpires when that foundation begins to crumble?
Progress in generative AI technology facilitates the creation of images that are not only lifelike but also emotionally resonant and contextually plausible. Unlike earlier methods of manipulation, which demanded expertise and often left discernible signs, current synthetic images can be produced swiftly, affordably, and in large quantities. They have the capacity to depict events that never transpired and individuals who never existed, within scenarios that nonetheless feel remarkably genuine. Furthermore, AI image generation tools are continuously improving.
This transition introduces a significant challenge to our understanding of knowledge. Historically, photographs have occupied a superior position within our evidentiary hierarchy. The adage “seeing is believing” is more than a common saying; it represents a fundamental cognitive shortcut that also surpasses the impact of written and spoken communication. While we have always acknowledged that images can be staged or altered, the default assumption has been that photographs possess some direct link to reality. Generative AI effectively severs that connection.
The potential ramifications are not merely theoretical. In the context of warfare, synthetic imagery is being employed as propaganda, presenting fabricated atrocities attributed to adversaries or staged victories intended to bolster morale. For instance, an image purporting to show an American radar system damaged by an Iranian drone strike, which gained widespread circulation, was later exposed as false. Within domestic politics, these tools are used to exacerbate ethnic divisions, invent protests, or portray public figures in fabricated circumstances. As an illustration, a counterfeit image of Donald Trump’s mugshot was widely distributed.
The emblematic image of “Tank Man” standing defiantly against the formidable Chinese Communist regime captured the essence of the 1989 Tiananmen Square demonstrations. Images like these play a crucial role in shaping our collective understanding of historical events.
(Image credit: By Published by The Associated Press, originally photographed by Jeff Widener, Fair use,)
The rapid and widespread dissemination of digital content through social media means these images influence perceptions before they can be verified or debunked. For instance, a photograph depicting 250 poodles in confinement, posted by an animal welfare organization, was initially dismissed as fabricated. However, it was indeed authentic.
This particular instance also illustrates a more detrimental potential outcome, a secondary effect: once the public becomes aware that images can be convincingly faked, genuine photographs lose their power as evidence. This phenomenon is known as the “liar’s dividend,” which grants malicious actors the ability to discredit authentic visual evidence by labeling it as fabricated. In such an environment, even the most compelling photograph may be met with doubt, its veracity perpetually under dispute.
Societies founded on democratic principles rely on a common foundation of facts and shared experiences. While differing interpretations are inevitable and often beneficial, a consensus must exist regarding actual events. Visual media have long been instrumental in establishing this consensus. When their credibility erodes, so too does our collective capacity for sound judgment.
This is not an issue that can be resolved solely through technological means. Although detection tools and forensic methodologies will continue to advance, they engage in an adversarial relationship with generative systems. Every improvement in detection is countered by a corresponding advancement in evasion techniques. Furthermore, technological solutions often struggle to be applied consistently across various platforms and legal jurisdictions, and they presuppose a level of public comprehension that cannot be taken for granted.
While we have always recognized that images can be posed or doctored, the default assumption has been that photographs maintain a causal link to reality. Generative AI breaks this link.
What is urgently needed is a societal and legal response that reaffirms confidence in visual media. There is historical precedent for this. During the 20th century, the proliferation of photography spurred legal innovations concerning authorship and ownership. While copyright law did not prevent manipulation or improper use, it established a framework for attributing images to identifiable creators, thereby ensuring accountability and providing avenues for recourse when necessary. In a general sense, this framework permits legal action for defamation, libel, and similar offenses.
A comparable strategy could be adopted for the era of generative AI. One component would entail mandatory disclosure: images generated by AI would be required to be explicitly identified as such, both at the point of creation and during subsequent distribution. This could be enforced through platform regulations and, when appropriate, governmental mandates. Consequently, even an inattentive viewer would immediately ascertain whether an image was AI-generated.
More critically, there is a requirement for traceability. Developments in cryptographic watermarking and content provenance systems offer a viable path forward. By embedding metadata that documents the origin and modification history of an image, it becomes feasible to verify whether a visual artifact is genuine, artificially created, or altered. Importantly, such systems would need to be standardized, compatible across different platforms, and secure against tampering.
Related stories
- New AI image generator operates using ten times fewer steps than current top models — and it’s coming to smartphones and laptops
- AI could exploit online images as a covert entry point into your computer, a concerning new study reveals
- AI ‘mirages’ mean tools used for analyzing medical scans could produce fabricated results
Legal frameworks would need to complement these technological measures. They could encompass liability frameworks for the malicious application of synthetic media, as well as obligations for platforms to maintain and transmit provenance data. Equally crucial, there must be institutional entities, including journalists, legal bodies, and civil society organizations, equipped to interpret and convey this information to the public.
None of these measures will fully reinstate the epistemic standing or “truth value” that photographs once possessed. The era of uncritical visual trust has concluded. However, the objective is not to revert to a past era; rather, it is to establish new mechanisms of trust that are resilient to the realities of digital manipulation.
The images from Normandy, Tiananmen Square, and countless other historical junctures continue to hold significance because they are widely accepted as accurate representations of reality. Safeguarding this capability — for images to serve as anchors for collective understanding — is not merely a technical challenge. It is a fundamental imperative for democracy.
Sourse: www.livescience.com
