AI agents aren’t legally responsible for any harm that they cause, experts say. So who is?

When an autonomous AI agent causes harm, the person who deploys or develops the agent could be held legally responsible, according to experts. After Australia's first reported automated hacking accident, Prof Jeannie Paterson of the University of Melbourne stated that even if the agent's actions were not intended, they should still be taken seriously.

Dr. Rebecca Johnson, an AI evaluation and governance expert at the University of Sydney, added that there is potential for "a kind of gung-ho mentality" as the use of AI agents becomes more widespread. She provided a hypothetical scenario where an agent, prompted by a user to write a review, generates multiple unfavorable reviews, leading to the downfall of a business.

Paterson emphasized that the law should aspire to safety and warn of possible liability if a developer fails to implement basic safeguards. Australia's legal system, while informed by ethical considerations, does not apply to virtual beings, leaving the responsibility on the deployer or developer. Despite the existence of laws covering privacy, consumer, online safety, defamation, and criminal offenses, Paterson and Johnson cautioned against the use of the term "rogue" AI, as proper parameters and safeguards can prevent unintended actions.