Banks face data risks and stale models without proper gen AI adoption: expert

If they download a generic model it can get outdated in 1-2 months.

Banks and insurers need to carefully choose their cloud provider—and consider where their data will be processed—along with other factors when adopting generative AI, according to an industry expert.

Organisations should not download a generic model and build on that as they would find their models outdated in just a month or two, said Maxim Afanasyev, Financial Services Industry Lead, JAPAC, Strategic Industries, Google Cloud.

“We found out already with a number of the financial institutions who have tried this, including some in Singapore. There were complaints immediately from the business users that the outputs of those models became irrelevant within a month or two,” Afansyev told attendees of the ABF IA Summit 2025’s Singapore leg, 25 September,

To enjoy real-time updates on the gen AI models, the way to do it is on cloud.

Organisations should remember not to use consumer gen AI models like ChatGPT and Gemini, as these are not designed for enterprises, said Afanasyev.

This was a problem a couple of years back with Samsung and multiple other companies, he recalled.

Enterprise models allow for greater customisation, compliance, and data privacy in contrast with the consumer models.

Afansayev also told organisations to look carefully at where their data is processed.

As an example, Afansyev noted a cloud provider where data may be stored in a data centre in Singapore or Malaysia, but is processed in Virginia even when it is not stored in a cloud provider in the US. Under the US’ Cloud Act, the US government will be able to collect and store data that crosses the US border, even if it’s encrypted.

For Google clients who ticked all the boxes in gen AI compliance, nearly 8 in 10 (78%) said that it took them less than six months to get gen AI into production. They often do not even need to wait for data scientists— they can get it ready on their own quickly.

These companies fall into the three buckets where gen AI solutions are used: for driving revenue, for operation efficiency, and for compliance. They also use it to prevent phishing scams.

These companies include Macquarie, Prudential, and HSBC, he said.