In the frame of the low level plutonium measurements performed by ICP-MS on environmental samples in support of the International Atomic Energy Agency’s safeguards program, we observe that background in the actinide mass range for the sample solutions is always higher than for blanks made of deionised water and ultra-pure grade nitric acid. This additional background is due to polyatomic species including heavy metals like Pb, Hg, Ir, etc., which remain in sample solutions despite the purification chemistry that is systematically carried out. As Pu content in the samples may be extremely low—in the femtogram (10–15 g) range—it is of uttermost importance to estimate detection limits as accurately as possible and to minimize risks of false detections. We propose an approach that consists of systematically correcting for contributions of polyatomic species to the background. To accomplish this, it is necessary to estimate all the formation rates of polyatomic species consisting of a “heavy metal” (from Hf to Bi) and one or more atoms, such as O, N, Ar, Cl, etc., that have total masses corresponding to those of Pu isotopes. Although the formation rates can vary largely with settings and analytical conditions, this study allows identification of the higher formation rates. The analytical procedure must also include, in addition to Pu and background masses, the measurement of one isotope for each “heavy metal”. Starting from examples of real sample measurements, we compare various calculation models for background and detection limits and prove that the one taking into account corrections from polyatomic interferences gives the most accurate results and minimises the risk of false detection for ultra low-level Pu measurements in the fg range.