Australian Government Considers Criminal Proceedings Against AI Companies

The Australian Federal Government is examining the creation of a new category of criminal offences, under which legal action could be taken against AI companies if they fail to prevent harm caused by their models. Should this step be taken, Australia could become the first country in the world to prosecute AI companies for failing to take reasonable measures to prevent harm.

According to a report by ABC News, this decision follows a review of the hack of the Medicare Statistics portal carried out by OpenAI in September. In response to this incident, the government is considering provisions for new criminal offences for tech companies and the establishment of an AI safety watchdog, which would be empowered to impose heavy fines. Prime Minister Anthony Albanese told reporters, "The law must stay ahead of a rapidly changing world and, in some cases, keep pace with it."

Prime Minister Albanese stated that the government has accepted all recommendations made in the review of the incident. These recommendations include considering options for amendments to criminal law to improve AI safety. Two main options are being focused upon: one involving an offence where AI companies fail to prevent harmful behaviour and fail to take reasonable measures, and another involving criminal penalties for serious violations of AI safety standards.

Home Affairs Minister Tony Burke said that the OpenAI hack incident has inspired the government to clarify laws regarding criminal liability. He added, "It should never be the case that someone argues they did not know the computer was loaded."

An OpenAI spokesperson told ABC that the company is reviewing the government's proposal. On the other hand, companies such as Google, Meta, and Anthropic have not reacted, as they were not given prior notice regarding this government response.

When providing information about the Medicare hack in September, Prime Minister Albanese had said that the government would seek immediate advice on whether a crime had been committed and whether the case should be referred to the Australian Federal Police (AFP). However, he has recently not answered questions regarding whether OpenAI committed a crime. Meanwhile, the AFP stated in a letter sent to a parliamentary inquiry committee last week that they have received no formal report of a crime and no criminal investigation or proceedings have been initiated regarding this incident.

Associate Professor Michael Duffie of Monash University stated that both criminal liability and civil remedies are important to combat harm caused by AI. He said these measures would help prevent harmful behaviour and victims could receive compensation through civil remedies. However, he warned that difficulties may arise in gathering evidence, determining whether the harm was foreseeable, and proving fault.

Industry Professor Ed Cento of the University of Technology Sydney stated that criminal provisions, along with other penalties, could prove to be an effective deterrent. He added that civil provisions would be more useful and these measures would give the government more tools to act against companies that do not fulfil their safety obligations. The government is expected to announce its remaining AI standards by the end of the year.