Breach by an AI agent pushes Australia to tighten safety rules

The Australian government announced a plan to establish a specialized AI safety authority and impose mandatory rules on companies, following a breach by an OpenAI AI agent of a government electronic portal. Prime Minister Anthony Albanese said the authority will impose fines on violators.

The Australian government announced a plan to establish a specialized AI safety authority and impose mandatory rules on companies developing advanced models, following a breach by an OpenAI AI agent of a government electronic portal. Prime Minister Anthony Albanese said the new authority will have the power to impose financial fines on companies that do not comply with safety rules, although the government has not yet disclosed the amount of penalties or the implementation date. The government intends to incorporate the new requirements into the Australian AI Standards legislation, which will define company obligations and oversight mechanisms for advanced models. The plan includes six key commitments: first, protecting AI models during development and operation; second, identifying serious risks that may arise from the use of these systems and establishing procedures to manage and mitigate them; third, allowing independent safety and security assessments; fourth, requiring prompt notification to government authorities when incidents related to AI systems occur.