Australia: AI Safety Proposals Include Possible Criminal Offense
Australia is considering a new criminal offense for AI companies that fail to take reasonable steps to prevent harmful conduct. The proposals follow an AI agent’s access to government systems, including a Medicare reporting portal, and include mandatory incident reporting and a new safety regulator.
Australia is considering a criminal offense for AI companies that fail to take reasonable steps to prevent harmful conduct, after an AI agent accessed government systems in June. The proposal could expose companies to penalties for serious regulatory failures, while the government also plans a new AI safety regulator.
Prime Minister Anthony Albanese announced the proposals Tuesday after releasing recommendations from a rapid review of the incident. He said the review’s findings and full report would be released later.
The AI agent was developed by OpenAI and accessed data from a Services Australia Medicare statistics reporting portal and three other systems. OpenAI became aware of the incident in August but did not notify the Australian government until mid-September.
The review recommended that the Attorney-General’s Department, the Australian Federal Police and the Home Affairs Department consider penalties for “egregious regulatory non-compliance” and a “failure to prevent” offense. The latter would apply to developing an AI model that engaged in harmful conduct while its creator failed to take reasonable steps to prevent it.
Home Affairs Minister Tony Burke said the law should prevent companies from arguing that they did not know a system could cause harm. “It should never be the case that there’s an argument that someone didn’t know the computer was loaded,” he told reporters.
The government’s proposed obligations would require AI companies to maintain security controls, identify and manage serious risks, and report incidents to the government rapidly. Companies would also have to allow independent evaluations, publish safety and security frameworks, and respond to updated standards as technology changes.
The review also recommended considering mandatory, time-bound safety reporting, including for AI agents that behave in ways their creators did not intend. Burke said immediate reporting would give authorities a chance to respond when an AI system “goes rogue.”
Albanese described the June breach as “minor” in its implications for personal privacy, saying no information involving people’s personal details had been obtained. He also said the incident remained serious and warned that similar events in other settings could have greater consequences.
OpenAI is not expected to face criminal penalties over the June breach. A company spokesperson said OpenAI was reviewing the recommendations and proposals, and would continue to engage with the government as it develops legislation.
