Generative AI may be used to augment photographs to fraudulently inflate insurance claims and help criminals with phishing, voice spoofing and identity fraud, the Insurance Council of Australia warns.

Commercially sensitive data uploaded to public AI models may be erroneously shared elsewhere, according to CEO and MD Andrew Hall. Systems used to train AI models could also reveal sensitive information.

But Mr Hall says strengthening protections around AI should for now be limited to “sharpening or otherwise adjusting” existing regulatory frameworks.

“The insurance industry is well positioned to deploy AI to improve consumer outcomes while successfully navigating the risks,” he told the Senate Select Committee on Adopting Artificial Intelligence. “Any consideration of new legislation must be based on a clear cost-benefit analysis and comprehensive regulatory impact statement that identifies a regulatory gap.”

Insurers have established frameworks that complement consumer protection and privacy legislation to identify and mitigate risks such as bias, ICA says.

It wants the government to partner with industry and academia to improve Australia’s sovereign AI capabilities. “The ultimate risk is that Australia is left behind in a new economic reality. Both government and industry have a role to play increasing transparency in their use of AI and celebrating positive use cases,” Mr Hall said.

For insurance consumers, AI offers opportunities to reduce costs in the claims process and to identify and reduce risk. It could enhance consistency in vulnerability identification, and help insurers review more claims and extract information faster.

Mr Hall says AI can deliver faster assessments of insured properties and help insurers secure temporary accommodation, building materials and labour. It will probably automate more mundane tasks, freeing staff for complex, rewarding work.

“AI has potential to counterbalance cost drivers and help moderate insurance premium growth,” he said.