Australia weighs new criminal offence for AI harms by companies

In short: Australia is considering a new “failure to prevent” criminal offence for companies whose AI models cause harm if the company did not take reasonable steps to stop it.

What happened

Australia’s government is looking at a possible new crime for some AI developers. The idea is called a “failure to prevent” offence. In simple terms, it would treat a company like a business that did not lock its doors, if its AI system was used in a harmful way and the company had not taken reasonable precautions.

This is not a law yet. It is a recommendation for further consideration that came out of a rapid government review.

That review followed a June cyber incident involving an OpenAI agent and a Services Australia Medicare statistics portal. The review called for stronger AI safety and security steps, including requiring companies to report serious incidents. The government said it accepted all 18 recommendations from the review.

Under the proposed approach, the Attorney-General’s Department would consult with the Australian Federal Police, the Department of Home Affairs, and other agencies about possible changes to criminal law. The review also raised the option of criminal penalties for the most serious cases of breaking AI related rules.

Why it matters

If Australia creates a “failure to prevent” offence, it could increase legal risk for companies that build AI models used by the public. For regular people, the goal is straightforward. It is to push companies to build stronger safeguards, so AI tools are less likely to be misused or cause damage, especially when they interact with government services and sensitive information.

Source: NYTimes