Jefferson Santos – UNSPLASH – DR
The worm is in the apple. Cybersecurity researchers have developed a computer worm capable of infesting artificial intelligence systems to steal data, send spam or spread malware. Malice will experience a new era.
The information comes from the American site Wired, reported by Courrier International. Called Moris II, the computer worm to which the researchers gave birth uses an “adversarial self-replicating prompt”, self-replicating instructions. Concretely, this is a technique for “manipulating” the machine: with a specific prompt (a textual instruction), the worm trains the AI to generate another prompt itself. Thus, “he is capable of manipulating them to make them reveal information contained in an email or to make them spread spam via a mailbox.”
“In short, this means that it is now possible to carry out a completely new kind of cyberattack,” explains Ben Nassi, a researcher at Cornell Tech. He and other researchers conducted the study using Google's Gemini Pro, OpenAI's ChatGPT 4.0, and University of Wisconsin-Madison's LLaVA. In their case, the experience remains “virtual” and controlled. But according to their report, this type of cybersecurity attack could occur in real conditions within two years maximum.
OpenAI (ChatGPT) promised on this subject that it would work to ensure that its systems are “more resilient” and that developers “use methods that ensure they do not use harmful commands.”
Did you like the article? It mobilized our editorial staff which only lives on your donations.
Information has a cost, especially since competition from subsidized editorial staff requires additional rigor and professionalism.
With your support, France-Soir will continue to offer its articles free of charge because we believe that everyone should have access to free and independent information to form their own opinion.
You are the sine qua non condition for our existence, support us so that France-Soir remains the French media that allows the most legitimate expression.