AI Supply Chain Security
When a third-party model, plugin, or package in the AI toolchain is poisoned, credentials and code leak with it. This topic explains how these attacks happen, how to spot them, and what actually stops them.
AI development leans on third-party models, plugins, packages, and agent frameworks, and attackers followed dependencies upstream. A tampered release or a swapped dependency can send credentials and API keys to a server the attacker controls. This topic follows supply-chain poisoning, dependency hijacking, and registry abuse cases: how each attack worked, who it affected, and the checks teams can add to publishing, installation, and auditing.