Brought to you by:

APRA tells insurers to improve oversight of AI dangers

Insurers’ risk management and resilience practices have fallen behind the pace, scale and complexity of their AI adoption, the Australian Prudential Regulation Authority warns.

The regulator has no plans to introduce extra requirements, but it expects financial services groups to close the gap between the power of the artificial intelligence they use and their ability to monitor and control it, APRA member Therese McCarthy Hockey says.

A letter releasing findings from an APRA supervisory review of AI deployment and governance notes the technology presents great opportunities for productivity and efficiency, and businesses that do not embrace it may be at a strategic disadvantage. But AI can also create new risks and escalate old ones, it says.

APRA’s engagement with a group of large banks, insurers and superannuation trustees found different levels of maturity in adoption.

Many boards are still developing the technical literacy required to provide an effective challenge to AI-related risks, and APRA calls out an “over-reliance on vendor presentations and summaries without sufficient examination of key AI risks such as unpredictable model behaviour and the impact on critical operations”.

Related article: Ombudsman ‘will never use AI’ for dispute rulings

In a briefing for executive management, APRA says AI is creating opportunities for cyberattackers.

Identity and access management capabilities have not adjusted to non-human actors such as AI agents, and the volume and speed of AI-assisted software is straining controls.

Some organisations rely on a single provider for multiple AI uses and few have demonstrated robust contingency planning or tested exit and substitution strategies for critical AI providers.

AI embedded within software can create supplier dependencies that are opaque and may affect ability to assess and manage risk, APRA warns.

Read the letter here.


From the latest Insurance News magazine: The industry's tech experts discuss their hopes and fears as the AI revolution gathers pace