Euro banks worry AI will increase their dependence on US big tech

Putting such a dominant power in the middle of your supply chain a risky move...

European banks are concerned that growing use of AI will serve to increase their dependence on the big US tech companies and lead to fresh risks for the industry.

The AI craze has seen organizations of all kinds rush to implement the technology, or at least evaluate how it might be employed to deliver an advantage in their line of business.

Yet the huge amount of compute resources needed to develop and train AI means that organizations risk becoming dependent upon the small number of companies that have control of these resources, which for the most part are US-based, a situation that has some banks worried.

According to Reuters these concerns were voiced at the recent Money 20/20 financial technology conference in Amsterdam.

Depending on a small number of tech companies is one of the biggest risks banks face, ING chief analytics officer Bahadir Yilmaz is reported as saying. European banks in particular need to ensure they can move between different tech providers and avoid vendor lock-in.

However, Yilmaz admitted that he expects to rely on the big tech companies more in future for infrastructure services, precisely because of the huge resources required for training AI models.

"You will always need them because sometimes the machine power that is needed for these technologies is huge. It's also not really feasible for a bank to build this tech," he conceded.

Joanne Hannaford, CIO of Deutsche Bank's corporate bank, agreed, saying: "AI requires huge amounts of compute and really the only way that you're going to be able to access that compute sensibly is from Big Tech."

Adrian Bradley, Head of Cloud Transformation at KPMG UK, told The Register: "It's important to note that banks' cloud requirements for AI may change over time due to the type and complexity of future AI projects. While the larger hyperscalers are often sought for their ability to support the training of large generative AI models, it is possible to run smaller models on local machines.

"While speed and accuracy may be lower, the benefit of running an AI model locally is the additional security and control over data. As such, banks should be mindful about where and how they deploy AI. This merely highlights the need for choice in the cloud market to drive innovation, reduce barriers to entry, and enable ease of adoption."

The hyperscale companies have been pumping investment into AI over the past year or two. Reports from market research companies show that the big cloud operators have been sucking up supplies of Nvidia's GPU accelerators in order to build out their AI services, likely expecting to make a profit from selling those AI services to customers such as the banks.

Many of the leading AI developers such as OpenAI are also US operations, and that particular company has a close relationship with cloud giant Microsoft.

However, US cloud companies are already facing antitrust investigations in the UK and Europe about whether they have altogether too much clout in the IT services marketplace; Brit regulator Ofcom found that AWS and Azure together accounted for 70-80 percent of the UK cloud infrastructure services market in 2022, with Google Cloud making up another five to 10 percent, for example.

Meanwhile, the UK's competition watchdog the Competition and Markets Authority (CMA) has also been examining the ties that have grown up between the big tech companies and AI startups, as The Register reported back in April.

The CMA warned that a handful of dominant technology firms – Google, Amazon, Microsoft, Meta and Apple (GAMMA) – could effectively shut down AI market competition through a web of partnerships, investments, and agreements.

A recent report published by the European Central Bank looked at the risks and benefits the financial industry faced from AI. It warned that while AI can potentially enhance the processing and generation of data, it may be prone to significant data quality issues.

In particular, the way foundation models are trained means that they may be more likely to acquire and incorporate biases or errors inherent in the data they have been trained on, and they are prone to hallucination the ECB said.

The report also picked out the risk of greater dependence on tech providers.

"Furthermore, depending on whether financial institutions have the in-house capacity to develop foundation models, the base architecture may need to be acquired from external companies. This will increase third-party reliance and could also raise data privacy concerns if the models provided by third parties are fine-tuned using confidential internal data," it stated.

At the end of May, the EU's financial markets regulator, the European Securities and Markets Authority (ESMA), issued guidance to financial services businesses using AI. This basically affirmed that uses of AI are covered by the EU's MiFID regulations, meaning that companies are still responsible for any actions taken.

Perhaps this is why Omdia Principal Fintech Analyst Philip Benton reports that the focus on AI at Money 20/20 was on the importance of keeping humans in the loop rather than replacing them.

"Some fintechs have been vocal about AI replacing the need for humans in certain roles, but the general sentiment at the event was about AI supporting the productivity of humans," he said. ®

More about


Send us news

Other stories you might like