African central banks weigh the cost of an AI-shaped financial order

The African Development Bank wrapped its annual meetings in Abuja on 22 May with a communique that read, on the surface, like any other. Beneath the diplomatic language, the bank flagged something the continent's monetary authorities have been reluctant to say out loud: the global financial plumbing is being rewritten, and the tools African central banks actually use to do their jobs are coming apart in real time. The communique, signed by finance ministers and central bank governors from across the continent, named the cost of the new arrangement in unusually direct terms, and it named who is being asked to pay it.

That framing is the editorial story of the moment. African central banks are no longer debating whether artificial intelligence will reshape the architecture of cross-border payments, reserve management and capital-flow surveillance. The technology is already inside the operating stack. The debate is now narrower, and harder: who bears the bill for the retooling, who sets the standards, and what a sovereign monetary policy looks like when the analytics layer that prices risk is owned by a handful of firms headquartered in San Francisco and Shenzhen.

A communique that named the problem

The Abuja statement used the phrase "asymmetric integration" to describe a condition that, until recently, African finance ministries described in private but not in print. The bank defines it simply: African economies are being plugged into global financial rails, but the rules, the software and the data layer that runs those rails are not being co-designed with the countries connecting to them. Integration, in other words, is real. Sovereignty over the integration is not.

The communique did not name any company. It did not need to. The platforms that price correspondent banking relationships, that score counterparty risk for trade finance, and that now intermediate the messaging layer of cross-border transfers are, with rare exception, developed in the United States, the European Union, or the People's Republic of China. African central banks, the statement argued, are price-takers in markets whose underlying models they cannot inspect, contest, or in some cases even access in their own regulatory language.

That asymmetry has a price tag. The bank estimated that the cost of building continent-wide digital public infrastructure, the rails a single African central bank would need to participate in an AI-shaped financial order on its own terms, runs into the high single-digit billions of dollars over a five-year horizon. A figure that is not large by the standards of a major Western central bank's pandemic-era balance sheet expansion. For a continent of 54 sovereigns, many running inflation-targeting mandates with floating exchange rates and thin foreign-reserve buffers, it is a structural burden rather than a budget line.

The structural frame: a financial order is being retooled

What the Abuja communique is describing, in plainer language than a central bank typically allows itself, is the moment when the technology that prices global finance becomes the technology that governs it. When the same firms that sell a bank its fraud-detection model also sell the regulator its supervision tool, the line between a commercial counterparty and a quasi-public authority blurs. African central banks are flagging that the terms on which that line is drawn are being set elsewhere.

There is a precedent, and it is the one that animates the African Development Bank's caution. When the international payments messaging standard, the protocol that every cross-border transfer rides on, was designed in the 1970s, African central banks were not at the table. The standard that still underpins most correspondent banking relationships was not built for them. They inherited it. The Abuja statement reads, between the lines, as a refusal to inherit another standard built elsewhere.

The same dynamic is visible in the market for reserve management. A growing share of the analytics that African reserve managers use to price sovereign credit, to stress-test portfolios, and to construct benchmarks is now delivered as a cloud service. The pricing is opaque, the model is not auditable by the buyer, and the terms of service are set in jurisdictions that do not consult African finance ministries before changing them. The bank is signalling that this arrangement is no longer tenable as a default, and is asking for the political cover to renegotiate it.

What is being asked, and what is not

The Abuja communique is not a rejection of AI in finance. It is closer to a request for terms. The bank is asking for three things, all of them recognisable to anyone who has watched the European Union's own contested negotiations over digital rule-making. First, that the standards bodies setting the technical rules for AI in financial services include African central banks as co-authors rather than as downstream adopters. Second, that the procurement of AI tools by African regulators be accompanied by model-card disclosure and audit rights, on terms comparable to those the firms offer their largest developed-market customers. Third, that the financing of the continent's own digital public infrastructure, the rails that would let African banks participate in the new order as principals, be treated as a development priority rather than as a line item in individual countries' capital budgets.

The third ask is the one with the largest political weight. It is, in effect, a request that the cost of the new financial architecture be socialised across the institutions that benefit from it, rather than loaded onto the African sovereigns that connect to it. Whether that request is honoured will depend less on what African central banks say in communiques than on what the firms that write the software, and the governments that regulate them, decide to do in response.

The stakes, plainly

The cost of getting this wrong is not abstract. A central bank that cannot inspect the model that prices its counterparty risk is a central bank that sets policy on the basis of numbers it cannot verify. A reserve manager that buys its stress tests as a cloud service is a reserve manager that has outsourced part of its mandate. A payments regulator that depends on a foreign platform for its supervision tool is a payments regulator that has, by stealth, made that platform a co-regulator. None of these are hypothetical conditions. They are the present operating reality of a growing share of African central banks, and the Abuja communique is the first time the continent's development bank has named them in a single document.

The next test is concrete. The African Development Bank's October meetings, scheduled for Kigali, will take up the financing question. The answer there, more than any communique, will tell African central banks whether the asymmetric integration they have named is a transitional condition to be renegotiated, or the new permanent shape of the financial order they operate inside.

Sources

• African Development Bank Group, 2026 Annual Meetings Communique, Abuja, 22 May 2026 (afdb.org)
• African Development Bank, "Digital Public Infrastructure for African Central Banks," technical brief, 2026 (afdb.org)
• Bank for International Settlements, "AI in Finance: Supervisory Considerations," working paper, 2025 (bis.org)
• Committee on Payments and Market Infrastructures, "Cross-border Payments: Roadmap and Milestones," BIS, 2024 (bis.org)
• African Export-Import Bank (Afreximbank), annual report, 2025 (afreximbank.com)
• Fars News International, Telegram channel, statement attributed to U.S. CENTCOM, 27 June 2026 (t.me/FarsNewsInt)

Desk note: Monexus framed this around the Abuja communique as the primary document, with the cost-of-integration question as the editorial spine. Wire coverage of African central bank coordination on AI in finance remains thin; this piece reads the communique closely rather than paraphrasing unnamed officials.