Backdoored AI code completions need forensic analysis as poisoned training data becomes a supply-chain risk

A Help Net Security piece describes a forensic tool designed to trace backdoored code completions generated by AI assistants. The concern is that code used to train large models ma