Robot farms, artificial intelligence and badbots are at the core of online campaigns to combat disinformation.

bots, are robots controlled by states or private interest that pose on the Internet as human beings to try and manipulate people during elections or political crises. In this case we are talking about Badbots. “bad robots” that can be difficult to detect, hiding behind false accounts and e-mails. They spread fake information on the Internet and engage with actual Internet users in order to destabilize countries like France.

franceinfo has penetrated Viginum for the very first time. Viginum is the service of the French state that fights against foreign interference in digital communications. Viginum alerts are delivered directly to the prime minister and intelligence agencies. Datalab analysts at Viginum are glued to screens to identify all social network “inauthenticity” – those activities that have been multiplied by bots, and increasingly by artificial intelligence. For example, we have an account which publishes all content at midnight or hourly, explains Paul Antoine Chevalier, the Datalab’s head. It’s not what a person would naturally do. It could be a bot or online AI. Most likely, it’s a machine. ”

AI in the service of manipulation and disinformation

Viginum was able to identify one of these automated disinformation operations against France during the Paris Olympics 2024: The pro-Russian Matriochka campaigns massively distributed fake content and fake answers for X by “bot farms”. Artificial intelligence is now helping to create a more convincing narrative. Paul-Antoine Chevalier confirms that it’s evident we are seeing AI manipulations when we perform our daily detections. Our in-house system detects massively duplicated contents, even when they are reformulated using large language models. The semantic distance between the contents can be seen. We need to know how to identify AI attacks, as it offers new possibilities to attackers. ”

It is claimed that Badbots have a significant share in the digital world. The American company Imperva (a subsidiary of French Thales) has been claiming to measure global bot traffic on networks for several years. Imperva claims that bots are mathematically on their way to overtaking human activity. When you examine all Internet traffic worldwide, half is made up of human activity and the remaining 50% are automated robots. Florian Malecki is one Imperva manager. Publishers in the cyber-security world and social networking providers need to do more to stop these badbots. But, like any crime, there is always a race to the finish between ‘good players’ and ‘bad players’. ”

“Today, 32% of bot traffic is malicious bot traffic. And this will continue to rise.”

Florian Malecki

to franceinfo

The public is polluted by these campaigns, amplified bots. The American presidential election, the war in Ukraine and Israeli-Palestinian conflict as well as legislative elections in France are all examples. Many specialists have documented the automation of disinformation or hate campaigns, including CNRS mathematician David Chavalarias.

He used a database of over 700,000,000 messages from 2016 to 2024 in a report to identify communities of pro Russian badbots. Bots have moved upmarket, and are now able to produce more human interaction than ever before. The bots have become more sophisticated. ChatGPT or conversational intelligences are now well-known to everyone. We know we can make robots which will contact users and dialog with them. They also analyze digital outputs and adapt their discourse. It is an actual danger. ”

The social network X is accused in the United States of having helped Donald Trump to win. X is constantly removing fake accounts to fight bots. Elon Musk’s platform has a vested interest in retaining many bots, as their activities bring traffic and lots money to advertisers.

OpenAI-Powered Automation Solutions