Australia Rejects Corporate AI Guardrails Amid Hacking Warnings
The Australian government rejected calls from AI firms to dictate safety rules while intelligence officials warned that outdated systems are vulnerable to AI-driven hacking.
Director General of the Australian Signals Directorate Abigail Bradshaw warned at the Sydney Dialogue summit in Canberra that Australia's outdated technology is vulnerable to AI-driven hacking attacks. Bradshaw noted that while the number of active AI agents on the internet is likely high, quantifying them is difficult and updating national systems will require enormous spending.
In response to the evolving threat, executives from OpenAI and Anthropic called for urgent global rule-setting and independent interrogation of AI models. However, Assistant Technology Minister Andrew Charlton rejected the notion of allowing private companies or the United States government to dictate AI guardrails. The Albanese government plans to release its own AI guidelines early next year.
Legislative tensions persist as Labor senators rejected a Greens' motion for mandatory safety testing. Additionally, the government remains unwilling to reduce copyright protections for AI training, despite requests from OpenAI and Anthropic. From the opposition, Coalition industry spokesperson Andrew Hastie urged a formal deal with the U.S. to attract AI hyperscalers to ensure Australia does not become a supplicant state.