According to what was reported by CrowdStrike on its annual report regarding cyber threats some groups linked to cybercrimine they would be abusing generative Artificial Intelligence for their own purposes.
Specifically, Il LLM model known as Llama 2 and proposed by Meta it would be a tool already used by cyber criminals.
According to cybersecurity experts, the language model was exploited to generate script for the task automation program PowerShell Of Microsoft. Through it, the cybercriminals would then download the access credentials of a computer system relating to unspecified individuals North American financial services.
To make the situation even more critical there is a consideration from CrowdStrike. According to analyses, in fact, the current ability to identify attacks of this type (therefore enhanced by AI) is quite limited. The documented cases are currently limited and similar phenomena could soon take on very high dimensions and levels of danger.
CrowdStrike is pessimistic about the near future”The continued development of AI will undoubtedly increase its potential misuse“.
The dark side of AI: not only abuses by cybercriminals but also misinformation among the risks
While cyberattacks are currently limited by the cost of developing LLM models, the latest technologies make potential cybercriminals’ work easier.
During 2023, there have been many attempts to exploit AI for cybercrime but, due to a technology that is still “immature”, they have often proved to be failures. This, however, could soon change.
In addition to cybercrime, there are other concrete threats related to AI. For experts, even the disinformation It is growing. In this sense, the 2024 it will be a notable test, with the 42% of the world population called to the polls.
As underlined by CrowdStrike, this is not science fiction but a reality already experienced in the context of Taiwan presidential electionswith China using AI-generated content to influence voting.