I made Michael_G.R.'s argument at the time, and despite even EY's claims, I don't think it violates the spirit or the letter of the rules. Remember, the question it's probing is whether a smart enough being could come up with a convincing argument you could not anticipate, and the suggestion that the gatekeeper consider the social impact of hearing the results is exactly such an argument, as others have indicated

Considering how hard it is for me to pin down exactly what the keeper has to gain under the rules from letting the AI out, I wouldn't be surprised if EY did some variant of this.