The accurate and rapid identification of gravitational wave candidates is crucial for follow-up observations of electromagnetic emissions. In run O3, the LIGO/Virgo collaboration released 80 Open Public Alerts. However, the rates of retractions and false alarms point to the need for a complementary model that predicts the nature and origin of OPAs in low latency. I will discuss GWSkyNet, a machine-learning algorithm capable of discerning between signals of astrophysical and terrestrial origin.