Recent EUCAST advice asserts that, with low breakpoints, susceptibility results for cephalosporins and carbapenems can be reported 'as found', even for strains with extended-spectrum β-lactamases (ESBLs) and carbapenemases. The CLSI has similar advice, but with higher ceftazidime and cefepime breakpoints than those of EUCAST. Pharmacodynamic and animal data are used to support these views, along with some analysis of clinical case series. We contend that such advice is misguided on three counts. First, whilst there are cases on record where cephalosporins and carbapenems have proved effective against infections due to low-MIC ESBL producers and low-MIC carbapenemase producers, respectively, there are similar numbers of cases where such therapy has failed. Second, routine susceptibility testing is less precise than in research analyses, meaning that ESBL and carbapenemase producers with 'real' MICs of 1-8 mg/L will oscillate between susceptibility categories according to who tests them and how. Third, although EUCAST continues to advocate ESBL and carbapenemase detection for epidemiological purposes, the likely consequence of not seeking these enzymes for treatment purposes is that some laboratories will not seek them at all, leading to a loss of critical infection control information. In short, it is prudent to continue to seek ESBLs and carbapenemases directly and, where they are found, generally to avoid substrate drugs as therapy.