Cybersecurity has become a field as transversal as AI is. If Artificial Intelligence is placing itself at the. Center of the information of everything that carries the IT label, it is not surprising that. The are redoubling their efforts to use it and, not only that, but to infect it. Recently, researchers at Cornell University have revealed a significant (and worrying) advance in this eternal game of cat and mouse, which is the creation of the first computer virus capable of spreading through artificial intelligence (AI) systems, dubbed ” Morris II. What is Morris II? Morris II is not a common virus.

Inspired by the very famous

approximately 10% of all computers that were connected to the Internet, this new virus represents a milestone in the history of cybersecurity related specifically to IAS, being the first to specifically target one of these systems, specifically LLMs. Researched and developed Malaysia Telegram Number Data team at Cornell University , Morris II has the ability to infect agents , particularly those designed to create content based on data and instructions they receive. Surely , Gemini, Llama, Dall-E … etc. come to mind. How does work? The virus uses what are known as “self-replicating prompts” to infiltrate systems. These prompts are basically sets of instructions that prompt the AI ​​system to generate and propagate more malicious instructions on its own.

To try to draw a parallel

we can think of old known attack techniques such as SQL injection or buffer overflow attacks. During experiments at Cornell, Morris II demonstrated its ability to attack autonomous messaging assistants, stealing information and sending harmful spam. In the same way, you can act on the generation of images from text. “Morris II” or the India Telegram Number AI ​​crook Scheme of how Morris II acts in an image generation prompt shown by the researchers. image © Cornell Tech. Marching one of Comparisons The self-replicating nature of Morris II is reminiscent of traditional computer virusesHowever, its focus on AI systems and the use of AI itself to generate its own attack paths are a separate chapter in the book of cyber threats.