Pandemic preparedness starts with an early warning system of viruses with a pandemic potential. Based on information collected in a multitude of surveys, hazard models were developed identifying influenza viruses presenting a pandemic threat. Scores are attributed for 10 viral traits by expert panels which identified avian influenza viruses (AIV) belonging to subtypes H7N9 and H5N1 as representing the greatest pandemic risk. In 2013, more than 100 human cases infected with AIV H7N9 were observed in China. Case fatality rate (CFR) was high (27%), but the human-to-human transmission rate was low and by serological evidence H7N9 did not spread widely. Nevertheless, until 2019 more than 1500 H7N9 patients were identified characterized by a high CFR of 39%. Serology demonstrated that mild infections with H7N9 were widespread. In 2003, more than 400 people experienced AIV H7N7 cross-infection causing mainly conjunctivitis during a large poultry epidemic in The Netherlands. Between 1996 and 2019, a total of 881 human infections with avian H5N1 viruses were documented showing a CFR of 52%. Outbreaks were centred on South East Asia and showed close associations with epizootics in poultry. Mutations predisposing to human cross-infections were identified in the haemagglutinin (HA) and the RNA polymerase subunit PB2 of human H7N9 isolates. Human H5N1 isolates showed mutations in the receptor binding domain of HA and transmission in mammals could be obtained by as few as four additional aa changes introduced experimentally. Researchers have defined viral point mutations in HA, PB2 and the nucleoprotein NP that allowed AIV to cross the species barrier to mammals with respect to receptor recognition, RNA replication and escape from innate immunity respectively. Based on this insight a sequence-based early warning system for AIV preadapted to human transmission could be envisioned. Mink farms and live poultry markets are prime targets for such sequencing efforts.