Banks Reduce Reliance on OpenAI Models, Analysis Reveals

Banks are reducing their reliance on OpenAI as their primary large language model provider, according to new data from AI benchmarking and intelligence platform Evident.

Eighteen months ago, OpenAI provided the underlying technology for roughly half of the disclosed AI use cases across the world’s largest banks. By the end of 2025, that share had fallen to around one-third, with financial institutions increasingly turning to alternative providers, including Anthropic and Google.

The shift comes despite banks being among the earliest and most enthusiastic adopters of generative AI. Since the technology’s rapid rise three years ago, the sector has acted as a bellwether for enterprise AI adoption, meaning changes in bank behaviour are often seen as an early indicator of broader market trends.

Data from Evident suggests OpenAI’s market share is being eroded as rival model providers gain traction. OpenAI has previously framed this trend as part of a wider move towards model-agnostic strategies, but bankers cited by Evident point to intensifying competition between AI labs.

According to those interviewed, OpenAI is no longer the only provider delivering models that meet banks’ operational and regulatory requirements, particularly as competitors accelerate product development and strengthen enterprise partnerships.

Coding has emerged as a key battleground. Widely regarded as banking’s first generative AI “killer use case”, it remains one of the few areas where more than half of deployed tools demonstrate tangible return on investment. In code generation and review, experts increasingly point to Anthropic as a leading provider, while Google has leveraged its existing cloud relationships to simplify integration.

That strategy was highlighted last month when BNY integrated Gemini Enterprise into its internal AI platform, Eliza, underscoring the growing role of alternative providers in banks’ AI stacks.