My personal hypothesis is that somebody read about passing a Turing test as proof of AGI and then proceeded to design a tool to pass said Turing test without ever even bothering to find out what AGI even is, let alone achieve it.