AI agents aren’t legally responsible for any harm that they cause, experts say. So who is?
TL;DR
After Australia's first reported automated hacking accident, legal experts warn that companies deploying AI agents may carry liability for what those agents do. Law professor Jeannie Paterson argues that a deployer can be responsible for resulting harm even without intent, as long as the harm was foreseeable. Developers could also be exposed depending on the circumstances. The account rests on a single source and leaves key details of the incident and its legal classification open.
Nauti's Take
The legal clarification is a gain for operators: once responsibility is named, it can actually be carried through scoped permissions, complete action logs and a human approval point. The weakness is the sourcing, since the report rests on a single incident and leaves the legal classification open.
Small teams should first check which harms their agent could realistically cause and whether every consequential action can be traced and stopped.
Sources