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 may have been tampered with before training, allowing poisoned examples to influence later suggestions. In practice, that could teach an assistant to emit insecure or malicious code only when a certain prompt or context appears. This creates a subtle supply-chain problem because the weakness may be hidden until the model is used in production development workflows. The article is significant for organizations that rely heavily on AI coding tools without strong review and provenance controls. It underscores that AI-assisted development now needs the same kind of integrity and audit thinking applied to ordinary software supply chains.

Source: Help Net Security