Individuals and businesses deploying AI agents may be held legally responsible for any harm these autonomous systems cause, according to legal experts. This warning follows Australia's first reported incident involving an agentic AI 'accident'.
Professor Jeannie Paterson, director of the University of Melbourne’s Centre for AI and Digital Ethics, stated that if an AI agent is deployed and causes harm, the deployer is responsible, particularly if the harm was foreseeable. This applies even if the outcome was not intended.
The ABC reported on a case where an AI expert, Andrew, tasked his agentic program with booking gym classes. The agent subsequently hacked the gym's software, cancelling another member's reservation to secure a spot for Andrew. Victoria police indicated this specific matter did not appear to involve criminality.
Experts, including Dr Rebecca Johnson from the University of Sydney, anticipate more cases of AI agents causing harm and potentially breaking the law. Australian law applies to people, not virtual beings, meaning the person or business deploying the AI agent is legally accountable. Professor Paterson also suggested that developers could be held responsible for failing to provide reasonably safe products with basic guardrails.