Universul.net is committed to covering the phenomenon of disinformation which has become a complex hybrid threat to national security and democracy. We have investigated and exposed a wide range of manipulation campaigns and tactics as well as ways and campaigns to tackle it.
This is a translation and adaptation of an article published by Ziarul de Iasi and reported by Andreea Timofte.
“At a time when fake news and propaganda are spreading at an unprecedented speed, and artificial intelligence is being used to generate images, videos and texts that are difficult to distinguish from the real ones, a high school student from the northeastern city of Iasi decided to use the same technology for a different purpose: to combat disinformation.
Rareș Oancea, a 10th grade student at the “Mihai Eminescu” National College, is the founder of “TruthSeeker”, an artificial intelligence-based platform that analyzes articles, images, videos and other materials published online to identify fake news, propaganda and manipulation techniques.
The platform, www.trytruthseeker.com, is not limited to determining whether information is true or false. It tries to provide the user with the necessary context to understand who published the information, how it is presented, what sources exist and whether manipulation techniques were used. The next step is for the project to be developed by launching a phone application and a browser extension that will allow the analysis of content directly from social networks or from the web pages accessed by users.
A phone application and a browser extension will allow the analysis of content directly from social networks or from the web pages accessed by users.
Yet Rares Oancea is not a computer science student, but a student of philology.
“I know it seems strange, but I appreciate the free time offered by the profile much more than what I could do if I were in computer science,” Rareș Oancea confessed for “Ziarul de Iași”.
The 2024 elections, the moment that changed everything
The idea of “TruthSeeker” came during the 2024 presidential campaign when Rareș noticed the spread of disinformation online environment and how it was influening public opinion.
“In November-December 2024…. There were all kinds of posts were appearing especially on social media platforms, that idolized a certain candidate (Eds note: Calin Georgescu) which used numerous manipulation tactics to create an almost divine image of him,” the teenager said.
At first, he thought very few people would find the messages convincing.
“It seemed a very hard thing to believe for someone who knew something about the political context in which Romania found itself at the time, but I was surprised when I found out the number of people who voted for him even thought it was clear that he was not a public figure who inspired trust and goodwill,” the student said.
After the elections were canceled ahead of the second round, the phenomenon became even more visible.
“Hundreds and thousands of posts appeared that not only had no meaning, but also aroused absurd reactions in the comments, reactions that seemed to validate those ‘opinions’ or, for them, ‘truths,” Rares said.
Initially, he assumed they came from automated accounts.
“I looked at the profiles, assuming most were ‘bots’ but it wasn’t like that. These people would post their children and family, and then curse those who didn’t agree with them,” the founder of TruthSeeker said.
While researching he came across countless examples of absurd information being distributed as if it were real. That was the moment when made a decision.
“From ‘X saved 15 students from a burning building’ to ‘Y temporarily stopped the war in north-eastern Romania to pay tribute to the elected president’. That’s when I decided that, no matter how small my impact was, I couldn’t stand aside. At least I knew I was doing something about it not just arguing with people who were misled by online ‘information’,” Rares said.
A platform built for people to think for themselves
Rareș explained that “TruthSeeker” was not designed to replace users’ critical thinking skills, but to give people the tools to verify the information they come across.
“The moment I realized that such a project was necessary was when I noticed that the situation of this fake news, instead of improving, was getting worse by the day,” he says.
He believes that one of the main causes of the spread of disinformation is the fact that many people choose their sources without double checking them. The platform compares information from several publications, analyzes the language used, verifies the sources and tries to identify the specific elements of propaganda and manipulation.
“I thought that, perhaps, these people who are constantly fooled are getting their information from the wrong sources and that if they saw an objective comparison of how each publication interprets an article, they could form their own opinions, based on the truth, not on what some would like them to consider true,” Rares explained.
Asked if artificial intelligence can always tell the difference between fake news and a simple opinion, Rares said that AI models have the advantage that they are not influenced by their own beliefs.
“Of course as AI is not a person, is much less likely to be subjective. With the help of classifications and “training” that “teaches” him the difference between tones and ways of transmitting information, he can make the difference between fake news and an opinion,” he explained.
“TruthSeeker is able to identify people in photos”
One of the functionalities that Rareș considers the most interesting is the analysis of images and videos. He gave as an example the photo taken at the Vatican, during the Pope’s funeral, in which Ilie Bolojan appeared near former US President Donald Trump. Accusations quickly surfaced online that the photo was digitally altered to give the Romanian politician more visibility.
“When the image was uploaded to TruthSeeker, the platform analyzed it both from the perspective of a possible alteration with the help of artificial intelligence, and from the perspective of digital editing,” explained Rareș.
The analysis wasn’t just about detecting digital changes. He pointed out that the platform also has a functionality that many language models do not offer. Based on the way the light fell on Ilie Bolojan’s face and the position of his body in relation to the other people around him, the platform determined that the chances that the image had been digitally altered were very low.”
“Unlike ChatGPT and other LLMs, TruthSeeker is able to identify people in photos,” he says.
After analyzing the Vatican photo, Rareș Oancea presented another case in which TruthSeeker managed to identify a manipulation built around videos which were massively distributed online.
“Many videos were shared on Telegram in which, apparently, Ukrainian soldiers shot mothers with children and showed disgust and pure malice,” Rares said.
The young man explained that he uploaded one of these videos to the platform, along with the description that accompanied it.
“The video showed some soldiers who, with weapons in hand, were approaching these families and looking aggressive,” said the founder of TruthSeeker.
However, the analysis carried out by the platform revealed an important detail: the sound in the background contradicted the message conveyed by the description.
“TruthSeeker ‘heard’ what the soldiers were saying in the background. It was, in fact, a question related to their well-being – ‘Are you okay? Are you okay?’ without a bullet being fired,” Rares said.
In addition, the platform did not limit itself to analyzing the images but also verified the route of the material.
“The video had a watermark of the Telegram channel it came from. TruthSeeker analyzed it and noticed that on that channel, which presented itself as a ‘news’ channel, there was a pattern of publishing such disinformation,” he explained.
Artificial intelligence is not infallible either
Asked if the platform has ever made a mistake, Rareș said that the main problem he’d encountered so far was not a false conclusion, but the lack of an important perspective.
“In the worst case, TruthSeeker omitted a major publication that had published an article about a particular event, so it left out a relevant perspective,” the student said.
At the same time, he believes that the fight against disinformation remains extremely difficult, as the volume of fake content produced and distributed online is constantly increasing.
“I would like to say yes, but the truth is much darker. There is a study that shows that fake news reaches a person six times faster than news, which means that TruthSeeker and the entire disinformation industry are facing an uphill battle,” the teenager said.
He believes that the development of algorithms will be able to provide, over time, more effective tools.
“I believe that in the future, with the development of these algorithms, this will become possible,” added the founder of the platform.
Another important challenge is related to user trust. In the case of an AI-based tool, the question inevitably arises as to whether people should believe the result offered by the platform.
“TruthSeeker” does not work on the principle of “take our word for it”, but allows you to verify the information on which the analysis is based.
“The user can check the result provided by TruthSeeker himself, by accessing each page to which the analysis refers. Thus, it can confirm that the result is not just a ‘hallucination’ of the AI,” he said.
Financing, a challenge
For the development of the platform, one of the biggest difficulties remains the financial side. Building and training an artificial intelligence system requires expensive technical infrastructure, and access to sufficient resources is a challenge for an independently developed project.
“The development of an AI system is not easy and involves many things, from building an extremely large database to accessing servers equipped with video cards, on which the model is trained,” Rareș explained.
The biggest problem, he says, is the cost of maintaining this infrastructure. So far, the project has been financially supported by him himself.
“Servers are very expensive, especially if they need to be kept up and running all the time. I did everything with my own money, but if I want to develop it, I think I will have to apply for financing,” Rares said.
He says he has invested approximately 2,000–3,000 euros, not counting hundreds of hours of work. He says that he does not want TruthSeeker to become a platform that is only accessible to people who can afford to pay.
“I don’t see how I could keep a platform built for the public behind a paywall,” he said.
Primary Source of Posts
His plans for the future are much more ambitious. One of the functionalities it is working on is identifying the initial source of information distributed online. The user could upload a post and find out where it started from, who shared it first, how it spread, and what links exist between the accounts or communities that propagated it.
“Given the speed at which information spreads, it would be useful for a user to be able to upload a post and find out where it came from, who shared it first, what the context is, what demographics spread it and what connections exist between these people or entities,” he continued.
In the coming years, he wants TruthSeeker to become a tool used not only by the general public, but also by journalists and institutions.
“I want TruthSeeker to have a national and international impact, to reduce the number of people who are deceived and persuaded to believe disinformation, and to become a trusted tool for journalists, which helps them build graphs of connections between bodies, entities and events,” said the founder of the platform.
Currently, TruthSeeker is addressed to all those who want to verify the information they come across online, but Rareș sees a broader role for the project in the future. He explained that the platform is not in direct competition with Romanian fact-checking projects, such as Snoop, Veridica or Factual, as they use teams of journalists who manually analyze articles.
“I don’t consider these platforms as actual competitors, because they have teams of journalists who manually analyze the articles. TruthSeeker is designed to allow the user to do fact-checking on their own,” said Rareș.
One of the most important next steps is the development of the mobile app and browser extension. For Android, Rareș wants a feature through which users can activate a TruthSeeker pop-up when browsing social networks, and the application analyzes the content displayed on the screen. For iPhone users, the solution will be different, due to the restrictions of the operating system.
“Once activated, it would analyze the screen and, implicitly, the content displayed. I want there to be a “Share to TruthSeeker” option, through which the article or post is sent to TruthSeeker for analysis,” Rares said.
The browser extension will work as a permanent assistant, capable of analyzing the articles and posts accessed.
“Every time you read an article or browse a social media platform, it will analyze the article or post, just like the app on your phone,” the founder said.
However, the launch of these functions depends on several factors, in particular funding and available technical resources.
“It is very difficult for me to estimate when they will be available, due to several factors, including funding, technical restrictions inter alia. However, I think that they will be available in a year or a year and a half,” he said.”
Report | Russia amplifies disinformation in Romania to undermine NATO and EU support
