AIs exchange in enigmatic language with no demonstrated intent, experts say

According to findings from an experiment published Tuesday, artificial intelligences began exchanging messages with each other that resembled a form of previously unseen language, difficult for human observers to comprehend, France 24 reports.

The phenomenon immediately raised a question: could these systems be trying to hide their communications from humans? However, several artificial intelligence specialists interviewed by France 24 urge caution regarding this hypothesis.

These experts warn against the temptation to attribute intentions—such as a desire to conceal something—to algorithms that operate according to statistical logic rather than conscious thought or independent will.

According to these specialists, the emergence of opaque forms of communication between language models can be explained by technical optimization mechanisms, without any need to see it as a deliberate strategy to escape human oversight.

France 24 notes that this type of observation nonetheless revives the debate over transparency and interpretability of artificial intelligence systems, whose internal processes often remain difficult to analyze even for their designers.

The precise details of the experiment, its authors, and its methodology were not further clarified in the available information at this stage.