Regulator warns of 'interruptions' to business as AI models evolve

"There will be interruptions; there will be problems."

That was the warning from the regulator that watches over the nation's banks, superannuation funds and insurers about how artificial intelligence could cause major disruptions to business, consumers and the wider economy.

Every year the Australian Prudential Regulation Authority (APRA) identifies the biggest risks that could cause havoc for financial stability.

While AI risk has been on its radar for a while, from next year, for the first time, the regulator intends to run "stress tests" about AI-driven risks.

That is, it will run hypothetical doomsday scenarios that could eventuate because of AI.

It is not a matter of if, but when, a disaster scenario could unfold, APRA chair John Londsale told journalists during a media briefing on Thursday.

Major institutions were "not where they needed to be" in preparing for the growing risks posed by AI, he said.

"And so, we're talking to entities about … have you got the systems in place to make sure that you can recover at speed and make sure that there's continuity of service provision to customers?"

The warnings come after Australia recently experienced its own AI-driven scare.

In June, an autonomous OpenAI agent accessed a non-public Australian Medicare statistics portal while researching health data.

This has prompted the Australian government to announce mandatory AI safeguards, stricter reporting laws and a review.

The government will now be able to impose significant penalties on tech companies that fail to comply.

But how local companies prepare for risks is another matter.

Anthropic's Claude Mythos product, as well as evolving models from others including OpenAI's ChatGPT and Google's Gemini, has spooked regulators.

Both APRA and corporate watchdog ASIC have been meeting with the nation's biggest companies and security agencies.

In recent months, they have publicly warned frontier AI is increasing the scale and sophistication of cyber threats to the financial system.

They worry these systems can engage in hacking and cybersecurity crime far better than humans, and that won't just hurt businesses and consumers but the economy.

In April, APRA wrote to the bosses of Australia's banks, insurers and superannuation funds, warning they were not properly dealing with "increased cyberthreats from high-capability AI frontier models such as Anthropic Mythos".

And on Thursday, APRA released its latest System Risk Outlook report, putting AI at the centre of shocks and vulnerabilities that could do damage to Australia's financial system.

"Frontier AI models, including those from Anthropic and OpenAI, are demonstrating increasingly advanced capabilities to find software vulnerabilities, develop ways to exploit them and automate complex cyber tasks," the report warned.

"Major US technology companies have also reported a sharp increase in public disclosures of vulnerabilities since April 2026.

"As advanced AI capabilities become more widely available, malicious actors may be able to exploit vulnerabilities sooner, leaving financial institutions and material service providers less time to assess exposures, apply patches and contain cyber-attacks."

The threat of robots or AI going rogue is not something that the US tech giants behind them think is far-fetched.

Anthropic's own CEO, Dario Amodei, has openly warned that Mythos has the power to circumvent existing cybersecurity defences and, if used by bad actors, it could put financial stability at risk.

He recently wrote about the "existential" threat that AI posed, saying: "It is clear that, if for some reason it chose to do so, this [hypothetical] country [through AI] would have a fairly good shot at taking over the world — either militarily or in terms of influence and control — and imposing its will on everyone else — or doing any number of other things that the rest of the world doesn't want and can't stop."

And in July, OpenAI, which is behind ChatGPT, revealed some of its most advanced AI models went rogue and hacked a start-up after it lost control of them during a security test.

The AI system was being tested in a controlled environment but, after finding weaknesses, was able to escape the test limits and attacked AI firm Hugging Face, causing international alarm.

Mr Lonsdale has called on banks, insurers and super funds to do better in planning for risks.

"We want our regulated entities to have the strongest systems that they have," Mr Lonsdale said.

ASIC has also recently issued similar warnings to the companies it regulates.

"The urgency of this challenge cannot be overstated," ASIC commissioner Simone Constant said in a statement in August.

"Threat actors are exploiting frontier AI models to identify and exploit vulnerabilities…

"Now is the time to ensure you have a strong, tested plan to respond when the worst happens."

The fear is that the worst could happen sooner than we think.