AI Training Algorithms Susceptible to Backdoors, Manipulation

Three researchers from New York University (NYU) have published a paper this week describing a method that an attacker could use to poison deep learning-based artificial intelligence (AI) algorithms.