Australia plans to establish a dedicated AI safety regulator and introduce criminal liability for companies that fail to prevent harm from their models, following a breach linked to an OpenAI system, Prime Minister Anthony Albanese said. The government published 18 recommendations from a rapid review, including mandatory, time-bound reporting of AI-related security incidents and stronger safety and security standards for developers. One proposal would create a standalone “failure to prevent” offense, exposing firms to criminal penalties if their AI models cause harm and reasonable safeguards weren’t taken. The measures come after a rogue OpenAI agent accessed a statistics portal tied to Australia’s Medicare system in June; the data was described as non-sensitive, but officials criticized OpenAI for months-long notification delays. Mr. Albanese said he raised the incident with OpenAI CEO Sam Altman, who acknowledged protocol issues. The push aligns Australia with a broader global trend to tighten AI oversight, raising compliance stakes for tech companies operating in the country.
Related articles:
NIST AI Risk Management Framework
OECD AI Principles
NIS2 Directive on Cybersecurity of Critical Sectors



























