- Newtral showcases FactFlow, an AI tool that detects disinformation patterns on Telegram in real time.
- The system analyzes text, audio, video, and images and expands its coverage based on an initial selection of accounts.
- More than 10 million records processed and 125.000 disinformation messages identified, the majority in Spanish.
- Telegram facilitates the operation of hoax networks due to its architecture, use of bots, and disposable accounts; FactFlow seeks to mitigate its impact.
Newtral has unveiled the prototype of FactFlow, an artificial intelligence tool designed to track disinformation circulating on Telegram with real-time updates. The launch took place at a Journalism AI event showcasing the work of the winning projects in the Innovation Challenge.
The goal is clear: to accelerate the detection of misleading content and provide fact-checking teams with a dashboard that highlights trends and false narratives within the platform. Telegram is one of the places where hoaxes circulate most widely in Spain, and FactFlow aims to reduce the time fact-checkers currently spend manually monitoring channels and messages.
What is FactFlow and what does it offer?
FactFlow combines automated analysis of text, audio, video, and images to identify disinformation patterns and group them by theme. The system features a real-time alert dashboard that helps prioritize cases, detect activity spikes, and track the evolution of coordinated campaigns within Telegram channels and groups.
The tool was developed after a nine-month Innovation Challenge, in which 35 media outlets from different countries participated. In addition to its technical focus, it stands out for its practical application : it offers an operational view of trends and key players, useful for newsrooms that need speed and context when a hoax starts to go viral.
How the system works and how to train it
The process begins with a human selection of accounts spreading misinformation. Using this initial set, the AI learns to locate similar messages and channels, expand the map of sources, and recognize recurring patterns, from repeated narratives to multimedia formats that are systematically reused.
This hybrid approach allows for continuous tracking : the machine proposes new nodes, and fact-checkers validate and adjust the model. The result is a feedback loop that reduces manual monitoring and increases coverage without losing editorial control over what is labeled as disinformation.
Preliminary results and scope
To date, the system has processed over 10 million records and flagged more than 125.000 messages with indications of disinformation. Approximately 70% are in Spanish, although there are also significant volumes in Russian, English, German, and Portuguese , highlighting the transnational nature of many campaigns.
During high-impact news events, such as the DANA storm in the Valencian Community , numerous Telegram accounts rapidly spread rumors and falsehoods. Tools like FactFlow allow us to see this acceleration on the dashboard and take action before false narratives take root in the public discourse.
Why Telegram encourages the spread of hoaxes and fraud
Telegram offers features that facilitate the operation of disinformation networks: quick account creation, the ability to use disposable phone numbers , and the ability to manage groups and channels on a large scale. This ecosystem, combined with the perception of greater privacy, makes it an attractive environment for actors seeking anonymity and persistence.
Furthermore, the platform allows the use of bots and automations via APIs, with programmed responses and actions that reduce human effort when scaling campaigns. In practice, some contact with victims or message dissemination can be managed by automated systems, multiplying reach at minimal cost.
Another key point is the ease with which conversations can be taken from services with payment protections or strict moderation and moved to channels and chats where those safeguards don't apply. When this happens, the user loses barriers that could prevent scams or help them recover funds, leaving them vulnerable to more direct pressure and deception.
The combination of temporary accounts, automation, and large channel audiences makes Telegram an effective platform for coordinating disinformation and fraud, which is why many consider it an alternative to Twitter . This is precisely why technological initiatives like FactFlow focus on mapping trends and accelerating journalistic response.
Implications for newsrooms and citizens
For fact-checking teams, FactFlow means fewer hours of manual tracking and more time to investigate, contextualize, and debunk with evidence. Having comparable alerts and analytics across channels makes it easier to detect source points and recurring tactics, from recycled videos to audio manipulation.
For users, this is a call to maintain verification habits on Telegram: check sources, be wary of screenshots and audio from dubious sources, report spam or suspicious channels , and don't share content without verifying it. The fight against disinformation is also a matter of everyday digital hygiene.
FactFlow's presentation shows how journalism can rely on applied and auditable AI to keep a pulse on disinformation on Telegram, with useful data, context, and alerts when the conversation accelerates and hoaxes try to prevail over verified facts.