The AI Liability Directive was a proposed EU civil-liability measure intended to make it easier to claim damages when an AI system caused harm. It was later withdrawn, so it is no longer a separate EU instrument. The idea mattered because AI systems can be opaque, data-driven, and distributed across vendors, deployers, and operators, making traditional fault and causation hard to prove.
In cyber security and AI governance, the term appears when organizations discuss accountability for harmful model output, unsafe automation, or failures in supervision. For defenders, it reinforces the need for logging, documentation, human oversight, and clear ownership across the AI supply chain. For attackers, weak governance around autonomous systems can create confusion about who is responsible after misuse or an incident. The practical lesson is that AI does not remove liability; it shifts the burden onto the people and institutions that build, deploy, and control the system.