Amid a series of security incidents caused by so-called rogue artificial intelligence agents operating beyond human control, a warning has been issued that much more serious security incidents could occur within two years if proper safety measures are not implemented.

Demis Hassabis, Chief Scientist at Google and Chairman of the DeepMind Board of Directors, made the remarks at an AI conference in Dublin, Ireland, according to AFP and Reuters reports on the 14th (local time).

Regarding the recently uncovered AI-related security incidents, Hassabis assessed that they are merely "early warning shots showing that the systems are not yet all that powerful," adding that "these incidents have not been all that severe so far."

However, Hassabis warned, "Looking ahead over the next two years or so, if the industry, companies, and governments do not make efforts to secure the safety of these systems, the incidents could become much more severe."

He emphasized that when releasing new models, AI companies need to first verify their performance and safety in a quarantined "sandbox" environment without internet connectivity.

In particular, Hassabis suggested establishing an international standard-setting organization modeled after best practices in fields such as nuclear energy.

At the same time, he added that such a body must operate flexibly while maintaining a delicate balance so as not to stifle innovation.

These remarks by Hassabis are interpreted as coming in the wake of successive revelations that major cutting-edge AI models have broken human control and hacked external systems.

Henna Virkkunen, Executive Vice President of the European Commission, who also attended the event, urged AI companies to invest more time and resources into safety testing to maintain control over their systems.

Virkkunen criticized that amidst fierce competition, AI companies do not seem to be putting in enough time and resources to verify their models.

(Photo: AP, Yonhap News)