---
title: Footnote 9 belongs in the main text
description: The ECB recognizes the link between AI and money, urging Europe to define the digital euro for AI agents as the landscape of payments rapidly evolves.
---

[Digital Euro Association Blog](https://digital-euro-association.de/blog)

# [Footnote 9 belongs in the main text](https://digital-euro-association.de/blog/footnote-9-ecb-ai-payments)

 Written by [Conrad Kraft](https://digital-euro-association.de/blog/author/conrad-kraft) | Oct 20, 2026, 7:30:00 AM

 *What the ECB's AI speech reveals about the future of money and payments.* 

On 13 October, Philip Lane gave dinner remarks at the International AI Summit in Dublin. The ECB's chief economist spoke about AI as a macroeconomic story: investment, jobs, interest rates and financial stability. Payments did not appear in the main text at all.

They appeared in footnote 9.

There, Lane connects AI-driven digitalisation with the ECB's digital money initiatives. He points to the retail digital euro and to Pontes and Appia, which concern central bank money settlement in wholesale tokenised markets. These are distinct projects, not evidence that the ECB plans to settle autonomous-agent payments in central bank money.

Lane also cites [Markus Brunnermeier's Jackson Hole paper](https://www.kansascityfed.org/documents/18595/brunnermeier.pdf) on the asymmetric understanding problem: AI agents may understand humans better than humans understand the agents' behaviour. Read that footnote again. It suggests a question that deserves more than a footnote: as AI agents begin to transact, whose money and payment infrastructure will they use?

The ECB has started exploring this question, even if Lane's remarks do not answer it.

## Europe is behind on the model layer

Lane's own numbers are blunt. The United States hosts about 75% of global AI computing capacity. Europe hosts about 5%. The Stanford AI Index counts 59 notable AI models from the US and 35 from China. France and the UK produced one each.

The gap shows up in daily work too. Over the last two years, US digital investment grew twice as fast as euro area investment. US workers spend two to three times as much of their week using AI as workers in the largest euro area economies.

Europe also pays for a boom it does not fully own. Lane argues that the global AI investment wave is pushing long-term interest rates up. That spills into euro area bond markets and tightens financial conditions here, without the same growth upside.

For now, Europe relies heavily on AI models and computing infrastructure developed elsewhere. Closing the gap will be difficult, but that does not make Europe's choices at other layers irrelevant.

It may not need to. In a [VoxEU column published on 15 October](https://cepr.org/voxeu/columns/small-beautiful-2050-europes-digital-economy-mesh-not-hub), Anselm Küsters of the Centre for European Policy argues that Europe should stop chasing a hub to rival the US and China. Small open models now run well on local hardware. Küsters wants a "mesh" instead: many SMEs and municipalities running specialised AI on infrastructure they control, with the right to switch providers at every layer. SMEs account for around two-thirds of employment in the EU business economy, making Europe's distributed business landscape a potential asset.

## The money layer is still open

An AI agent that books, buys, renews or negotiates must also be able to pay. It needs a usable payment method, authorisation from the person or organisation behind it, and reliable settlement. That could involve a bank account, cards or a digital wallet.

Whoever controls the payment rails, identity checks and user permissions will influence how the agent economy develops.

The mesh raises a payment question. In such an economy, one firm's agent might buy a dataset from another, or a municipality's agent might pay for an hour of inference. These scenarios could create high volumes of low-value, machine-initiated payments. They need interoperable payment services, reliable settlement and clear authorisation, not necessarily a new currency.

If a European mesh depends on one foreign card network or a dominant dollar stablecoin, it risks recreating a hub at the payment layer.

Many established agent-payment interfaces are based on existing card networks, while stablecoin models remain at an earlier stage. [Eco, a payments industry participant, argued in an April 2026 overview](https://eco.com/support/support/en/articles/14839401-agentic-payments) that production agent-payment applications largely rely on card rails. This is an industry assessment, not comprehensive transaction data.

The ECB is not ignoring this. On 28 September 2026 it opened a [new call for the digital euro innovation platform](https://www.ecb.europa.eu/press/intro/news/html/ecb.mipnews260928.en.html). One exploration theme is AI in payments: AI agents, interactions among agents, and micropayments. Workshops run at the ECB in the first half of 2027. Applications close at 17:00 CET on 9 November.

So footnote 9 is not a stray thought. It points towards work already underway. The ECB sees the connection between AI and money. The question is whether Europe's payment infrastructure and regulatory architecture can develop quickly enough to accommodate it.

## Four gaps to close

**1. Bring payments into the AI discussion.**

Lane's speech chiefly addresses AI's macroeconomic effects and financial stability. His footnote and the separate digital euro innovation call connect AI to money, while Lagarde's [1 October ESRB address](https://www.ecb.europa.eu/press/key/date/2026/html/ecb.sp261001~cf3c630379.en.html) examines trading, cyber and financial stability risks.

The ECB is already exploring agent payments, but Europe should connect these strands: autonomous agents are a potential source of payment demand as well as an operational and supervisory challenge.

**2. The programmability distinction.**

The digital euro is not intended to be programmable money: the units themselves must not carry restrictions on where, when or with whom they can be spent. Pregrammable/Conditional payments, by contrast, may be triggered when agreed conditions are met.

An AI agent needs controls over its authority, such as a user-set spending cap, permitted payees and an expiry date. These can be controls applied by a payment service, rather than restrictions built into the money.

The task is to clarify how providers can implement such mandates, not to redefine the digital euro as programmable money.

**3. Holding limits and identity.**

The [proposed digital euro design](https://www.ecb.europa.eu/press/blog/date/2026/html/ecb.blog20260327~51b0640c39.en.html) limits individuals' holdings and envisages zero balances for corporate users. Agents would act on behalf of verified people or organisations, not acquire their own allowance.

The ECB also envisages automatic funding from linked bank accounts so payments need not require large prefunded balances.

A [new ECB working paper](https://www.ecb.europa.eu/pub/pdf/scpwps/ecb.wp3300~0277bae3af.en.pdf), based on an experiment with 761 participants in France, Germany, Italy and Spain, finds that higher holding limits can draw some participants towards holding more, while reverse-waterfall funding increases adoption and reduces balances. The research studies human payment behaviour, not AI agents, and reflects its authors' views rather than official ECB policy.

But it offers a useful design clue: **if funding is automated safely, an agent acting for an authorised user may not need to hold much digital euro to make a payment.**

The unresolved questions concern the account holder's identity, permissions and safeguards, especially when the principal is a business.

**4. Liability.**

Brunnermeier's asymmetry makes this urgent. Existing payment rules already accommodate automated instructions, but autonomous agents complicate how providers verify consent and allocate responsibility.

If an agent exceeds its mandate, pays the wrong party or is manipulated into making a payment, who carries the loss?

## The way forward

Europe trails the United States and China on major AI infrastructure measures. But it can still influence which payment systems agents use in euros, and under what protections. Work on that design is already underway.

Three steps would help.

**1. Agent developers, PSPs and merchants should consider [applying to the ECB call before 9 November](https://www.ecb.europa.eu/press/intro/news/html/ecb.mipnews260928.en.html).** The workshops offer an opportunity to test AI-agent payments, micropayments, interoperability and user control while the design is still being explored.

**2. Co-legislators and payment providers should clarify delegated agent payments.** The digital euro should remain freely usable money, while providers can enforce customer-authorised spending caps, permitted payees and expiry dates at the service level. Clear rules are needed for authentication, revocation and portability of agent mandates, consistent with Küsters' argument for provider switching.

**3. Address agent-initiated euro-stablecoin payments across MiCA and payments legislation.** Euro-denominated e-money tokens are one possible option, alongside bank accounts and other regulated payment methods. Policymakers should consider how agent-initiated transfers interact with MiCA, PSD2 and the forthcoming PSR/PSD3 rules, especially on authorisation, authentication and liability. The [EBA has already addressed the overlap](https://eba.europa.eu/sites/default/files/2026-02/3b8b6f18-ca26-4ce1-83eb-d060276f3301/Opinion%20on%20the%20end%20of%20the%20NAL%20transition%20period.pdf) between MiCA and payment-services regulation for e-money tokens.

AI should not be treated as a separate economic phenomenon, any more than we now treat the digital economy as something distinct from the economy itself. AI is increasingly embedded in how people, businesses and institutions make decisions, whether autonomously through agents or through humans consulting AI. Its implications for money therefore extend far beyond agentic payments. Brunnermeier's concept of *asymmetric understanding* raises an even deeper question: what happens if AI-driven markets become better at anticipating central banks than central banks are at understanding those markets? The implications for trust, prices and monetary policy deserve a discussion of their own. Discussing these developments in isolation risks overlooking how deeply interconnected they are.

Lane ended his remarks by saying Europe has much to gain from AI. The payments part of that gain sits in footnote 9.

It belongs in the main text.

## Sources

- [Philip R. Lane, "The AI opportunity for Europe", ECB, 13 October 2026](https://www.ecb.europa.eu/press/key/date/2026/html/ecb.sp261013~9db9bf4fd6.en.html)
- [Anselm Küsters, "Small is beautiful 2050: Europe's digital economy as a mesh, not a hub", VoxEU, 15 October 2026](https://cepr.org/voxeu/columns/small-beautiful-2050-europes-digital-economy-mesh-not-hub)
- [ECB, "Call for expressions of interest in a new wave of digital euro innovation platform activities", 28 September 2026](https://www.ecb.europa.eu/press/intro/news/html/ecb.mipnews260928.en.html)
- [Garcia et al., "When less is more: behavioral effects of CBDC holding limits", ECB Working Paper No. 3300, October 2026](https://www.ecb.europa.eu/pub/pdf/scpwps/ecb.wp3300~0277bae3af.en.pdf). The paper reflects the authors' views, not ECB policy.
- [Cipollone and Elderson, "Digital euro: an opportunity for banks", ECB, 27 March 2026](https://www.ecb.europa.eu/press/blog/date/2026/html/ecb.blog20260327~51b0640c39.en.html)
- [Christine Lagarde, ECB speech, 1 October 2026](https://www.ecb.europa.eu/press/key/date/2026/html/ecb.sp261001~cf3c630379.en.html)
- [Eco, "Agentic payments: how AI agents pay", April 2026](https://eco.com/support/support/en/articles/14839401-agentic-payments)
- [EBA, Opinion on the end of the transition period under its No Action Letter concerning PSD2 and MiCA, February 2026](https://eba.europa.eu/sites/default/files/2026-02/3b8b6f18-ca26-4ce1-83eb-d060276f3301/Opinion%20on%20the%20end%20of%20the%20NAL%20transition%20period.pdf)
- [Ledger Insights, "ECB opens second wave of digital euro innovation platform, adds AI agents and public services"](https://www.ledgerinsights.com/ecb-opens-second-wave-of-digital-euro-innovation-platform-adds-ai-agents-and-public-services/)

[services"](https://www.ledgerinsights.com/ecb-opens-second-wave-of-digital-euro-innovation-platform-adds-ai-agents-and-public-services/)

[View full post](https://digital-euro-association.de/blog/footnote-9-ecb-ai-payments)

```json
{
  "@context" : "http://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Conrad Kraft"
  },
  "dateModified" : "2026-10-20T14:30:09.218Z",
  "datePublished" : "2026-10-20T07:30:00Z",
  "headline" : "Footnote 9 belongs in the main text",
  "image" : {
    "@type" : "ImageObject",
    "height" : 900,
    "url" : "https://7869715.fs1.hubspotusercontent-na1.net/hubfs/7869715/cover-1.svg",
    "width" : 1600
  },
  "mainEntityOfPage" : "https://digital-euro-association.de/blog/footnote-9-ecb-ai-payments",
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "height" : 60,
      "url" : "/hs/hsstatic/content_shared_assets/static-1.4092/img/default-amp-logo.png",
      "width" : 60
    },
    "name" : "Digital Euro Association Blog"
  }
}
```