In case they don't answer, I would guess they are referring to VAERS data, and some random guy who once claimed it underreports everything by 40x. This is not logistically plausible (where would you hide 40x more patients than expected arriving at hospitals?), and I have seen some recent studies suggest that the underreporting factor is actually <1x (meaning it is overreported) when doing statistically random sampling. But that does not stop people from repeating the claim ad nauseam.