The rating agency Moody’s has warned that the financial sector’s race to adopt AI is putting banks at the mercy of a small group of Silicon Valley firms, creating systemic risks around outages, vendor lock-in, and price control.
The Guardian’s report outlines a familiar tension: AI will eventually cut costs and increase revenues for banks, but the path there requires substantial investment, creates new vulnerabilities, and concentrates power in the hands of the few companies that build the models and run the cloud infrastructure.
The Core Warning
Moody’s report identifies a structural risk: “The reliance of most financial firms on a relatively small set of foundation AI model and cloud computing providers risks creating a systemic dependency.” The concern is straightforward. If one major AI provider suffers an outage, the impact could spread quickly across customers and sectors — not unlike the cloud outages that have already taken down banking services in recent years, but with more embedded dependencies.
The rating agency also flagged what it calls “vendor dependence risk” — the possibility that dominant AI providers could, over time, control the price of AI services. This is particularly relevant given that the biggest AI companies are loss-making. OpenAI and Anthropic are both under pressure to deliver returns to investors. When the path to profitability runs through the institutions that pay for your models, pricing power becomes a strategic question.
More than 75% of City of London companies now use AI, according to a UK Treasury select committee report published in January. The applications range from automating administrative tasks to processing insurance claims and assessing creditworthiness — core banking functions, not peripheral experiments.
What the Banks Are Actually Doing
Lloyds Banking Group’s chief executive Charlie Nunn recently outlined a £13 billion strategy that uses AI to lure new business, improve efficiency, and increase shareholder payouts. The plan includes £2 billion in cost cuts, which Nunn acknowledged will affect staff. “That is going to impact work,” he said. “It is going to require us to continue to reskill people, but that’s been my history for 30-odd years in financial services.”
Moody’s estimates there’s a 20% chance that by 2030, AI will be able to do the work of a “solid mid-level employee.” For a sector that employs hundreds of thousands of people in roles that are mid-level by definition — loan officers, claims processors, compliance analysts — that’s not an abstract concern.
The rating agency also raised the possibility that AI could make it easier for customers to switch accounts for better interest rates, creating the risk that large deposits could move at short notice. “In this context, depositors’ trust in the institution and the resilience and stability of deposit funding are critical,” Moody’s said.
The Counterweight
Moody’s isn’t blind to the other side. Banks aren’t passive consumers. Many have decades of experience negotiating tech contracts and the leverage to push back on pricing. Some are exploring open-source AI models and strategic partnerships to offset dependency. And critically, banks retain control over their proprietary data — the one asset the tech firms can’t replicate.
The rating agency also acknowledged that the benefits are real. AI adoption will eventually cut costs and grow revenues. The problem is that with every bank racing toward the same goal using the same small set of providers, many of those benefits will be “competed away” — meaning the gains flow to customers and shareholders rather than sticking to the banks themselves.
Why NZ Should Pay Attention
New Zealand’s big four banks — ANZ, ASB, BNZ, and Westpac — are subsidiaries of Australian parent companies that are themselves heavily invested in AI. The systemic dependency Moody’s describes applies here too. If a major AI provider suffers an outage or raises prices, the impact flows through the Australian parents to the NZ subsidiaries.
The Reserve Bank of New Zealand has been increasingly focused on operational resilience for financial institutions. The Moody’s report adds a specific dimension to that conversation: not just “can your systems handle an outage” but “how many of your critical functions depend on the same AI provider as every other bank.”
NZ’s smaller market means fewer alternative providers and less negotiating leverage. The vendor dependence risk Moody’s identifies is arguably more acute for smaller economies that buy AI services from the same handful of global providers as everyone else.
The Regulatory Gap
The Bank of England was handed powers in July to regulate key tech firms including Amazon and Google — a recognition that the financial system’s dependence on cloud providers has reached a point where the regulators need to oversee the providers, not just the banks. Moody’s report suggests that framework may need to extend to AI model providers specifically, not just the cloud layer underneath.
As AI adoption deepens, regulators may increase their focus on operational resilience and third-party concentration in the AI model stack. That’s Moody’s language, and it’s a signal that the rating agency sees this as a credit risk — something that could affect how banks are rated, not just how they operate.
