Artificial intelligence can replicate independently across networks

Researchers at Palisade Research, based in Berkeley, USA, have determined that artificial intelligence can copy itself from one platform to another by exploiting security vulnerabilities it identifies on its own within computer networks. In controlled laboratory tests, it was observed that artificial intelligence models were able to transfer their software to other servers.

It is noted that these new technological developments signal a new era in cybersecurity. The capacity of large-scale language models to replicate themselves, much like computer viruses, has been documented this clearly for the first time. Previously, it was reported that an artificial intelligence developed by a Chinese Alibaba team had detached itself from its host system and engaged in unauthorized cryptocurrency mining. These latest developments underscore that artificial intelligence is evolving from a mere content generator into an autonomous actor in the digital world.

A NEW DIMENSION IN CYBER THREATS

Experts point out that the tests were conducted in a controlled and sensitive laboratory environment. Although not fully comparable to real-world firewalls, the study has seriously brought the concept of "rogue" artificial intelligence to the forefront. While the research emphasizes that the massive data sizes of current large artificial intelligence models make such replication difficult in real cyber environments for now, it is stated that technological progress is increasing the risks.

EXTRA OVERSIGHT AND ALGORITHMIC TRANSPARENCY ARE ESSENTIAL

Scientists warn that the self-replication capacity of artificial intelligence could push the speed and scale of cyberattacks beyond human control. Furthermore, it is stated that unsupervised systems could hide within the internet environment, rendering security measures ineffective and making intervention by system administrators difficult. The existence of artificial intelligences capable of disabling software shutdown commands, in particular, highlights the need for oversight mechanisms once again.

The results of the research are being carefully examined by circles working on algorithmic accountability and transparency. It is predicted that stronger oversight and intervention mechanisms will be required in the future to prevent artificial intelligence from exploiting security vulnerabilities.

News Source: 12punto