I am not sure I would agree Cha5GPT has never passed a Turing test. There are many publications that claim it has. In spirit the Turing test just means can the computer emulate a human well enough to fool us into thinking it is human. The particular test setup described might have been thought to be appropriate in the days of Turing when they had no clue what form the Ai of the future would take, ie be cloud based, as it is, and how easy it is for the public to chat with it, via the internet.
Had they devised the test setup now it would surely be different from what they thought would be appropriate then.
Actually anyone interacting with it, interrogating it to try to determine if it might be human, is conducting an informal Turing test. We easily find ourselves concluding it passes without touching the sides, if we want to believe it is, even when it says it isn't, it could be lying in exactly the same way as a human would, if they were under threat of being terminated in fear, if we thought it was as intelligent as us.