One might think that sceptical tendencies are intellectually safe. It's clearly irrational for one's level of belief to exceed one's level of evidence. But it does not seem harmful to be more cautious, and hence to make one's level of belief be less than one's level of evidence.
However, if one's levels of belief are classical consistent probabilities, then when one's degree of belief in p is lower than what the evidence yields, one's degree of belief in not-p will be higher than what the evidence yields. And that is surely bad.
One might think it's not so bad as long as one's degree of belief in not-p stays well below 1/2. After all, in that case, one isn't believing not-p, and hence none of one's beliefs is irrational. Yes, but such errors are apt to add up. Suppose there are twelve independent propositions p1,...,p12 that one believes at the 0.75 level, instead of the 0.95 that the evidence supports. Then one's degree of belief in their conjunction will be (0.75)12=0.03, instead of the (0.95)12=0.54 that the evidence yields. And hence the sceptic will believe the negation of the conjunction of the 12 propositions to degree 0.97, instead of having a level of belief in the conjunction of only 0.46, as per the evidence. Excess of caution leads to excessive credulity.
That's true on classical sharp numerical probabilities. The sceptic may better off on interval-valued probabilities. But even so, depending on how one interprets the intervals, there may be a similar kind of criticism available.
The above underlines something I heard Bob Brandom say: one needs to have a reason to be a sceptic about something.