The Australian Prudential Regulation Authority has warned of a “rapid escalation” in AI-related risks, as it prepares to launch a stress test of the Australian financial system’s exposure.

“Australia’s financial system remains resilient, but the risk environment is becoming less forgiving,” APRA chair John Lonsdale said.

“It’ll be the first time that we do a stress test …  specifically focusing on AI, and that’s a very deliberate decision … risks to financial system stability are increasing, primarily due to growing geopolitical volatility and the rapid escalation of AI-related risks.”

APRA says next year’s stress test will examine how AI-driven economic disruption could affect banks, superannuation funds, life insurers and the broader financial system.

“The exercise will look at risks that could emerge in severe but plausible downside scenarios, either from rapid AI adoption disrupting some business models or a sharp downward reassessment of AI-related asset valuations,” the regulator says in its System Risk Outlook, released today.

The report is released twice a year and the latest edition shows APRA’s concerns, especially over AI, have risen sharply since the last evaluation in May.

“Frontier AI is amplifying cyber threats and creating new operational risks,” the report says. “Agents with autonomous capabilities could create additional risks if they pursue assigned goals through actions that are not authorised.

“For example, in June this year an OpenAI agent circumvented controls to access an Australian government website. Without effective human oversight and preventative controls, agents’ pursuit of their goals may lead to unintended harm.”

The regulator is also stepping up its work preparing for quantum computing.

“APRA is encouraging the banks, insurers and superannuation funds it regulates to begin identifying where vulnerable cryptography is used across their systems, data and material service providers. This includes critical infrastructure and sensitive information that must remain secure over long periods”

The report says AI-related investment is also changing the risks general insurers cover.

“For example, the scale and concentration of investment in data centres is testing conventional approaches to assessing and underwriting risk. Clusters of facilities and high-value equipment within individual centres mean a single event, such as a natural disaster, could cause substantial losses. The scale of some projects is also stretching available insurance capacity.”

See the report here.