Innovation. Risk. Gold Bars. A Fireside Chat with Sarah Breeden

We ended this year’s UK Finance Digital Innovation Summit on a high — from the hidden vaults underneath the Bank of England to the heart of financial stability.

In a wide-ranging and refreshingly candid fireside chat with Sarah Breeden, the Bank of England’s Deputy Governor for Financial Stability, we explored everything from the opportunities of AI and deep tech to the systemic risks that come with it. We even took a detour under Threadneedle Street, where 400,000 gold bars are safely stored in the vaults of the Bank — around £400 billion at today’s prices.

Stability as the foundation for innovation

Sarah kicked off by challenging the common assumption that innovation and stability are in tension. “They’re not opposing forces,” she said. “Stability enables trust, and trust enables innovation. If we want innovation to endure, we need to build it on a stable foundation.”

That theme — innovation with purpose and responsibility — echoed throughout the day and framed this final conversation.

We talked about artificial intelligence, deep tech, programmable money, even the application of machine learning. The benefits are huge. But so are the risks. As Sarah put it: “Our job isn’t to slow innovation. It’s to shape it — and make it safe.”

The AI shift: from assistive to agentic

AI is already embedded in financial services — in fraud analytics, risk management, and trading. But the next wave, generative and agentic AI, brings new dimensions of risk.

Sarah explained:

These models learn and evolve independently
Their objectives may not align with public interest
They can be opaque, unexplainable — even to those who deploy them

The Financial Policy Committee is watching closely. Among the top concerns:

Black box risk: Can firms really understand what the model is doing?
Market volatility: AI-driven herding or even collusion between AI agents could amplify instability
Third-party dependencies: Echoing concerns we’ve seen in cloud, but now with AI
AI-enhanced cyberattacks: Faster, harder to detect, and potentially more devastating

Sarah asked, “Would guidelines help?”. “Especially as we shift from AI assisting people to AI acting for them?” It’s a shift we’re already starting to see. And one that requires new thinking on governance, accountability and resilience.

Experimentation in a safe space

The Bank isn’t just observing — it’s acting. Sarah shared how it’s using AI internally, including to analyse 50,000 responses to the digital pound consultation. And through initiatives like Project Rosalind, the Digital Securities Sandbox, and its AI consortium, the Bank is working with industry to test and scale innovation safely.

These efforts are all part of what Sarah described as the Bank’s broader role: to build the infrastructure and frameworks that enable others to innovate with confidence. And central banks? They may not always be first movers, but they are — and must be — foundational enablers.

A final reflection

One of the most striking moments came when Sarah recalled a conversation with Reid Hoffman, co-founder of LinkedIn, who had visited the Bank for a previous session.

Asked about the risks of AI collusion in markets, Hoffman offered a sobering comparison: “The same features apply to autonomous drones used in military operations.”

It was a reminder that the risks we’re discussing are not theoretical. They’re real, and they’re already here.

But that’s not a reason to pause innovation — it’s a reason to get it right.

As Sarah put it in her closing words: “We want innovation. But we want it safely. That’s our job.”

Area of expertise:

Payments, innovation & resilience