AI authenticity risks undermine digital banking

Banks warned to secure data integrity amid rapid AI adoption.

Banks and fintechs are accelerating their use of AI-driven agentic systems, but a growing authenticity problem threatens the foundation of digital financial services. According to Roy Fielding, Senior Principal Scientist at Adobe, the industry still lacks the safeguards needed to ensure AI agents deliver verified, bank-sourced information to customers.

“The main challenge for agentic AI use across the internet is trust,” Fielding said on the sidelines of Singapore Fintech Festival 2025. As more consumer interactions are mediated by AI, he warned that banks risk losing control over how their information is represented and interpreted.

Traditionally, the web relies on mechanisms such as the Domain Name System and corporate website certificates, allowing users to authenticate the entities they communicate with. But this certainty collapses in the world of AI systems trained on massive, unvetted datasets. “The agents are trained using external data, not specifically the bank's data… We don't know where the data has come from, whether it's authoritative or not,” Fielding said.

Fielding also drew parallels between today’s AI boom and the explosive growth of the early web. “AI has been growing at a very fast rate, but roughly the same rate which the worldwide web grew at during the early 90s,” he noted. Back then, international forums played a critical role in establishing standards; now, he believes the industry must again collaborate—this time to secure content authenticity and ensure data used in AI and API interactions is verifiable.

On fears that AI sits atop a speculative bubble, Fielding remained measured. “All bubbles eventually burst,” he said, adding that only companies that “collaborate on safe ways doing business online” will endure.