IAS (Image Authentication Software)

In spring 2023, social media feeds were flooded with eerily realistic images: Vladimir Putin kneeling before Xi Jinping, an explosion at the Pentagon, Joe Biden playing in the rain. None of it had actually happened – these were AI-generated fakes.

With powerful image-generation tools now widely accessible, distinguishing real from artificial visuals is becoming increasingly difficult.

The problem? There is currently no standard, regulation, or technical safeguard that allows us to reliably distinguish between AI-generated images and real ones.
This is especially concerning because images have long been perceived as trustworthy sources of truth. Photographs, in particular, are regarded as factual representations of reality, hence the popular phrase, “pics or it didn’t happen.” Research has shown that people are more likely to believe a piece of information when it’s accompanied by an image. This deep-rooted trust in visual media has historically been exploited to manipulate public perception, such as when Stalin famously doctored photos of Lenin addressing a crowd, erasing individuals who had fallen out of political favor.

History has shown how easily visual media can be manipulated—now, AI makes that process faster, more convincing, and more available than ever. As images continue to play a central role in how we understand the world, the question of authenticity is no longer optional—it’s essential.

Authenticity through Metadata

AI-generated images, like real ones, contain Exif metadata. This software reads and visualizes that data and uses AI to fill in missing details. It helps assess whether an image has been altered or artificially created—even when information is incomplete.Designed as a native Mac app, the tool mimics a professional utility for verifying visual content before sharing, especially in public contexts. It highlights the growing need for careful handling of visual misinformation.
Until recently, no regulations required AI-generated images to be labeled. Users relied on visual clues or context—but these methods are no longer reliable as AI evolves. Inspired by the EU’s AI Act, the software applies its own metadata standards, including:

• Creation and modification time
• Software and ownership info
• Risk level (based on the AI Act)
  
• The original prompt used for generation

The software analyzes the available metadata for completeness, cross-checks individual data points, and identifies potential inconsistencies. In addition, users can perform their own integrated research, for example, by using map services or comparing GPS coordinates to verify the content of the image independently.