machine-learning

AI agents turned poisoned repositories and obfuscated code into supply-chain execution paths

AI agents changed repository poisoning from a code-review problem into an execution problem because models now read untrusted project text, interpret obfuscated code and act through trusted developer tooling.