IBM's AI watermarking method protects models from theft and sabotage

Researchers at IBM have developed a method to watermark machine learning models in order to protect against theft and adversarial attacks.