Australia's ageing government systems are a sitting target for AI-driven cyberattacks, spy chief warns

Key points

• Australia's signals intelligence chief, Abigail Bradshaw, has warned that old government technology leaves the country exposed to AI-powered cyberattacks.
• Replacing legacy infrastructure costs enormous sums, and agencies face pressure to keep systems running around the clock.
• Anthropic CEO Dario Amodei warned that AI bots could flood the internet within a year, calling for a slower pace of development.
• Sam Altman of OpenAI and Elon Musk have both expressed support for slowing AI development.
• Australia's government is currently negotiating rules to govern how AI can be used inside the country.

Abigail Bradshaw, head of one of Australia's leading signals intelligence agencies (organisations that monitor electronic communications for national security threats), has put a stark warning on the table: the country's old technology is wide open to attacks powered by artificial intelligence.

Her concern is practical. Replacing ageing government computer systems costs an enormous amount, she said, and those systems can't be taken offline for upgrades because the public now expects them to be available at all times. Old code you cannot easily patch, sitting on a live internet connection, is exactly the weakness that AI-assisted hacking exploits.

Australia is currently working out what guardrails, meaning rules about what AI can and cannot do, should apply as the technology develops rapidly. That policy conversation is happening in Canberra right now.

What are AI-powered cyberattacks, and why does old technology make them worse?

AI-powered cyberattacks use software that scans thousands of systems automatically, finds weak points faster than any human team, and adapts when defences block it. Old systems are more vulnerable because they were built before many modern security standards existed, and vendors often stop releasing security patches for them.

For ordinary Australians, the risk isn't abstract. Government systems hold medical records, tax details, social security data and benefit records. A successful attack on any of those could expose personal information or disrupt services people depend on daily.

What is Anthropic warning about, and who agrees?

Dario Amodei, chief executive of Anthropic, the company behind the Claude family of AI assistants, warned that AI bots, meaning automated software programs that act on their own, could swarm across the internet within roughly a year. He called publicly for the pace of AI development to slow. Sam Altman, who leads OpenAI (the organisation behind ChatGPT), and Elon Musk have both backed that position, as we reported on 13 September in our story on Altman and Musk's rare show of agreement.

These aren't outside critics flagging the speed as a problem. They're the people building the most capable systems in existence. That's what makes the warning harder to dismiss than a regulator's white paper.

What should ordinary Australians do?

You don't need to act on anything today, but knowing where your data sits helps. If a government service you use announces a security incident, update passwords linked to that account promptly and watch for unusual activity on financial accounts. That basic drill applies regardless of whether the cause is AI-assisted or not.

Common questions

Why is old technology so hard to replace?

Many government systems run software that is decades old and deeply woven into daily operations. Switching them out takes years of work and hundreds of millions of dollars, and agencies cannot afford downtime while the replacement is built.

Could AI attacks affect me directly?

If a government database holding your details were breached, your personal information could be stolen and used for fraud. Regulators in Australia require agencies to notify affected individuals after a confirmed data breach, so watching your inbox after any major incident is worthwhile.