Just to be sure, are you talking about
flame atomic absorption spectroscopy (FAAS)?
As a first step, make sure the instrument is well cleaned and
aligned. Then warm up the instrument for a while before you
analyze. After warm up, try running the samples. You may see
some drift. A small amount of drift is not uncommon.
Check your method. Some methods have you zero on the blank.
In that case, you may just need to rezero the instrument.
If neither of those work, the negative blank says you have a difference
in the matrix between the blank and the samples. A good approach
with AA is to use the same matrix for both samples and blanks. Otherwise
the difference in absorbance can translate into concentration problems and
interferences.
Hope this helps.