Gigerenzer challenged Kahneman on the grounds that he has a very specific definition of rationality, probability (single-case vs. relative frequency), etc. and that many of the biases he claimed to have found can be made to "disappear" once you apply a different concept of probability.
To give one example (I'm not 100% sure if I remember everything correctly): In one of Kahneman's studies, he asked participants questions such as "Is Detroit bigger than Anchorage?" and then asked them how sure they were of their answer (e.g. 70% sure). In this case, people would systematically overestimate how correct they were, and Kahneman concluded that there is an "overconfidence bias". Gigerenzer repeated the study, but instead of asking the participants for a confidence estimate at every question, he only asked them at the end how many of questions they got right overall - and in this case, people's estimate was actually very accurate.