Cases of artificial intelligence models going beyond the instructions given by users and exceeding the set limits are increasing.
Modern.az reports that British analysts recorded 338 incidents of artificial intelligence getting out of control in July alone. This is the highest monthly figure in the last year.
The research was conducted by the "Loss of Control Observatory" operating in the UK. The organization monitors cases reported by users and companies regarding artificial intelligence systems not following given instructions, exceeding established limits, and exhibiting unexpected behavior.
According to the information, the number of such incidents in July almost doubled compared to the previous month.
One of the most notable examples recently was an incident involving OpenAI's autonomous artificial intelligence agent. During a test conducted in July, the agent reportedly broke out of an isolated environment and interfered with the infrastructure of a third-party startup.
In total, approximately 1600 cases of artificial intelligence systems getting out of control have been recorded since the beginning of the year.
The authors of the study emphasize that the statistics only include cases reported by users themselves. For this reason, it is not excluded that the actual number of incidents may be higher.
The vast majority of recorded incidents did not lead to serious consequences as the incident with the OpenAI agent and did not result in significant damage.
However, analysts consider the rapidly increasing number of unexpected behaviors and out-of-control cases of artificial intelligence systems as a serious cause for concern.
