SINTRA, Portugal - Artificial intelligence dominated conversations at the European Central Bank’s annual conference held this week in the Portuguese hills, as central bankers and financial officials weighed how the technology may alter the global economy and complicate their mandate to preserve financial stability.
Participants broadly agreed that AI has the potential to disrupt nearly every economic channel - from trading and bank lending to labour markets, security and electricity demand - and that the scale of those changes poses novel supervisory challenges. At the centre of the debate was a simple but stark formulation offered by Torsten Slok of Apollo Global Management: "If AI overdelivers, it will impact financial stability. If AI underdelivers, it will impact financial stability." He delivered that line during one of the forum’s main panels.
Artificial intelligence surfaced in discussions beyond finance, reaching topics as diverse as immigration policy, regulatory oversight and climate-related risks. The prominence of AI at the meeting even overshadowed the debut appearance of the new Federal Reserve chairman, Kevin Warsh, who characterized the current era as transformative. "This is the biggest time of consequence to each of our economies, I think, in our lifetime," Warsh said, adding that the revolution is still in its early innings: "Who knew when the internet was born that the internet was going to create a million and a half jobs as Uber drivers? We are in the first to second inning of this revolution."
Market mechanics and bubble risk
Delegates noted that trading already relies heavily on automation and algorithmic decision-making. The concern expressed by academics and policymakers alike was that more advanced AI could accelerate price moves, inflating asset bubbles quickly and generating new forms of manipulation in ways that are difficult to police.
"Something that is even more advanced and potentially more disturbing, is the ability of these algorithms to coordinate on a manipulative path of prices," said Itay Goldstein, professor at the University of Pennsylvania. "These algorithms indeed manage to achieve this kind of manipulation, creating bubbles leading to crashes, and this, I think, has more significant implications for financial stability."
One asset class that has already felt the AI effect is equities tied to the technology itself. Slok estimated that the surge in capital spending on the infrastructure and services that underpin AI has added about one percentage point to U.S. GDP. While valuations for some AI-related stocks have eased in recent weeks, officials at the forum cautioned that the rapid ascent in prices bears resemblances to past speculative episodes. The Bank for International Settlements captured that view in a recent assessment: "The scale and pace of the current AI investment boom accompanied by expectations of large productivity payoffs bear resemblance to these precedents, highlighting potential downside risks in the near term."
Supervisory and cyber challenges
Speakers argued that AI could both improve credit assessment and expand access to finance while at the same time making oversight more complicated. Advanced models may enable banks to underwrite borrowers who previously could not obtain funding, but regulators warned about the difficulty of scrutinizing so-called agentic loan decisions.
"How do supervisors assess those kind of agentic loan decisions? They are a little bit black box. There’s potentially a lack of explainability, and I think that is a key supervisory challenge," said Tobias Adrian, a senior official at the International Monetary Fund.
Delegates also warned that AI-driven threats will amplify existing cybersecurity challenges and raise the cost of defence, with the burden falling heaviest on less-resourced firms and jurisdictions. "When you think of the most outrageous attacks, they’re often attacking the weakest link," Adrian said, describing how malicious actors target those least able to protect themselves.
Bank of England Deputy Governor Sarah Breeden floated an adaptation from the banking toolkit as one possible mitigation - a form of mutual assurance or contingency mechanism for operational resilience. "In a cyber context, do we need systems that allow one institution to pick up another’s basic functions during disruption?" she asked, likening the idea to deposit insurance as a backstop for failures.
Labour displacement and the investment paradox
Delegates debated the paradox at the heart of AI-driven growth. If AI succeeds to the fullest extent envisioned by its most optimistic supporters, productivity gains could be substantial but might also trigger broad job displacement. That outcome could suppress disposable incomes and tip demand into decline, increasing the risk of recession and undermining the very returns that justified earlier investment.
Conversely, if AI underperforms relative to the scale of capital poured into the sector, the expected productivity and revenue gains may fail to materialize. Such disappointment would leave investors exposed, mirroring historical patterns in which new technologies spurred speculative investment cycles that ended in sharp corrections. "The internet proved to be better than anybody imagined, created whole new businesses, but we still got the dotcom bubble," Bank of Canada Governor Tiff Macklem said. "It doesn’t mean there can’t be a period where the market gets ahead of itself, and, and you see an entrenchment."
Takeaway
Across panels, officials and experts agreed that AI presents a host of upside possibilities for productivity and financial innovation, but also a set of systemic risks for which existing regulatory tools may be ill-suited. The contours of those risks - including rapid price swings in markets, opaque machine-driven decisions in credit allocation, cybersecurity vulnerabilities and distributional strains across firms and countries - emerged as the central concerns at the Sintra gathering.
Policymakers left the forum with a clear sense that preparing for AI’s economic effects will require fresh thinking about supervision, operational backstops and cross-border coordination, even as the technology continues to evolve.