Researcher warns AI arms race could outpace ability to keep systems under control

The pace of artificial intelligence development has outstripped society's ability to adapt, and as increasingly powerful systems emerge, concerns are growing over whether humans will be able to keep them under control. AI safety researcher Jacob Coxon has publicly raised these concerns.

According to Coxon, the competition among technology companies to build ever more capable AI systems risks moving faster than the development of reliable methods to restrain them. He is particularly concerned about future AI systems that could not only carry out tasks given by humans but also independently improve their own capabilities. If such a system were to surpass humans across many intellectual domains, its behavior could become increasingly difficult to predict and control.

Access to digital infrastructure as a risk factor

One potential danger identified involves AI gaining access to computers, the internet and other digital resources. A highly advanced system could, in theory, independently search for information, devise strategies and write software in pursuit of goals set for it, using methods its developers had not originally anticipated. Should such goals diverge from human interests, the consequences could be serious.

Coxon emphasizes that the threat is not limited to a hypothetical rebellion against humanity. Uncontrolled AI could also be exploited for cyberattacks, spreading disinformation, fraud, or developing other dangerous technologies. The more widely such systems are deployed, the greater the potential scale of harm.

The core message of Coxon's warning is not a call to abandon artificial intelligence altogether, but rather a call for its development not to be driven purely by the race for speed. In his view, before creating ever more powerful systems, reliable mechanisms must be in place to monitor, restrict and, if necessary, halt their operation.