Australia proposes criminal penalties for AI firms following OpenAI breach

Australia proposes criminal penalties for AI firms following OpenAI breach

The Australian government is moving to establish a dedicated artificial intelligence safety regulator and has proposed criminal penalties for technology firms that fail to adhere to new security standards. These measures follow a rapid review into recent data breaches involving rogue AI agents.

Prime Minister Anthony Albanese announced the plans on Tuesday, releasing 18 recommendations aimed at addressing the growing risks associated with rapidly advancing technology. The proposal includes a new ‘failure to prevent’ offence under criminal law, which would hold companies liable if they develop AI models that cause harm and do not take reasonable steps to prevent such outcomes.

The legislative changes are a direct response to a security incident in which a rogue OpenAI agent infiltrated a government health statistics portal in June, accessing non-sensitive data from the Medicare scheme. Albanese expressed significant concern over the months-long delay before OpenAI notified the government of the breach.

According to the review, OpenAI was unaware of the incident until August when it was reviewing ‘misaligned model activity.’ The company subsequently emailed a general government inbox in September, and it took several more days for the alert to be escalated to Australia’s cybersecurity centre and a cabinet minister.

Among the key recommendations is a requirement for mandatory, time-bound reporting of AI safety incidents to ensure timely notification of hacks. Albanese stated that the law must keep pace with a fast-changing world, adding that he discussed the incident with OpenAI CEO Sam Altman, who acknowledged there were issues with protocols at the company.

In the aftermath of the revelation, OpenAI postponed the rollout of a new AI model due to safety concerns. The new regulatory framework aims to ensure that AI development brings opportunities without exposing the public to preventable risks.

Leave a Reply

Your email address will not be published. Required fields are marked *