NextFin News - Canadian Prime Minister Mark Carney is calling for a global "technology stability board" to oversee artificial intelligence, modeled on the Financial Stability Board that coordinated the world's response to the 2008 financial crisis, arguing that only international coordination can keep AI development safe without handing the rulebook to Washington or Beijing.
The proposal arrives at the exact moment the AI industry's own leaders are split over whether to slow down. Anthropic chief executive Dario Amodei published an essay on Saturday urging frontier-model developers to pace capability gains, and OpenAI's Sam Altman, SpaceX's Elon Musk and Google DeepMind chair Demis Hassabis quickly voiced support. U.S. President Donald Trump has dismissed the guardrails camp, saying the United States must stay ahead of China. Into that rift steps Carney with an institutional answer: not a slowdown, but a coordinator.
"There's a need for coordination," Carney said in an interview on Monday. "Ultimately, a technology stability board, in our view, along the lines of the Financial Stability Board would make sense."
He framed the move as pro-growth rather than restrictive: "It's an opportunity — and I think ultimately what it's going to mean is greater value creation and greater value delivery to people."
The FSB Playbook: Soft Power, Hard Consequences
To understand what Carney is proposing, it helps to remember what the Financial Stability Board actually does. Created at the G20 Pittsburgh summit in April 2009 as the successor to the G7-only Financial Stability Forum, the FSB was never given treaty powers. It cannot fine banks or block mergers. It sits in Basel, hosted and funded by the Bank for International Settlements, and brings together finance ministries, central banks and supervisors from all 20 major economies plus key financial centres such as Hong Kong, Singapore, Spain, Switzerland and the Netherlands, alongside the European Central Bank and the European Commission.
Its weapon is coordination. The FSB identifies vulnerabilities in the international financial system, issues recommendations to G20 leaders, runs peer reviews of national implementation, and — crucially — turns national regulatory preferences into de facto global standards through a process the industry then adopts to keep market access. Post-crisis bank capital rules, too-big-to-fail resolution regimes and derivatives reforms all flowed through that channel. The rules were not "global law"; they were G20 consensus that became domestic law in every jurisdiction that mattered.
That is precisely the model Carney is reaching for, and it is a deliberate choice. A UN-style treaty body would require ratification that the United States, in its current posture, would be unlikely to grant. A pure voluntary code would be ignored. The FSB template offers a third path: a members-only club of national authorities that coordinates quietly, publishes peer pressure, and lets the largest markets enforce the resulting norms through their own jurisdiction. It is governance by gravitational pull rather than by statute.
The historical precedent is stronger than it first appears. The FSB's predecessor, the Financial Stability Forum, was a G7 creation; the 2009 upgrade to the FSB deliberately pulled in emerging economies — Brazil, China, India, and others — precisely because a club that excluded the fastest-growing parts of the system could not govern it. Carney's pitch for a technology stability board carries the same logic: a body that excludes either the United States or China cannot govern AI, but a body that includes both needs the soft-law flexibility the FSB pioneered.
For Canada, the pitch is also about positioning. Carney's government is convening an investor summit in Toronto with more than 100 high-profile global investors and a prospectus of over 160 projects needing hundreds of billions of dollars. "Foreign capital is underallocated to Canada," he said, adding that his government would soon announce tax-competitiveness reforms and accelerate federal project approvals to one year. A seat at the table of a new global tech body is part of the same sales pitch: Canada as a stable, rules-based jurisdiction where AI capital can park itself between the American and Chinese poles.
The Timing: A Middle-Power Bid in a Two-Pole World
The immediate trigger is the Amodei essay, titled "We Must Pace the Frontier." In it, Amodei argued that the industry must slow the rate at which model capabilities improve so that risk prevention has time to keep up. "We must slow the pace at which we improve the capabilities of A.I. models," he wrote. "Progress will still seem fast, and we must make wise use of the time we gain." His proposal has three parts: embedded third-party evaluators with permanent access inside every frontier lab, coordinated pacing among democratic countries, and eventually global coordination that includes authoritarian governments. Anthropic is unilaterally committing to the first step.
His condition was telling: any slowdown must be coordinated, "without sacrificing commercial advantage or the United States' lead in AI." Even the industry's leading hawk on AI risk knows a unilateral pause is commercial suicide. That admission is the opening Carney is exploiting. If the industry itself concedes that safety requires coordination, then the question becomes: coordination under whose rules?
The United States, under Trump, has answered with export controls and chip leverage — a unilateral toolkit. The European Union has answered with the AI Act, whose enforcement machinery switched on August 2, 2026: the EU's AI Office can now demand technical documentation from general-purpose AI model providers, evaluate models, order corrective measures and impose fines. That is the world's first binding supranational AI regulator with actual teeth, and it began exercising real supervisory power over the frontier-model market barely six weeks before Carney's proposal.
Carney's proposal is a middle-power attempt to insert a multilateral layer between those two hardening blocs. It is consistent with the broader foreign-policy line he laid out at Davos in January 2026, when he urged the world's "middle powers" to combine their weight against great-power rivalry and the erosion of the rules-based order. It also builds on Canada's own domestic foundation: the AI for All strategy launched June 4, 2026, a five-year plan built around roughly C$2 billion in new federal investment, targeting US$200 billion in additional economic growth, 250,000 new AI-related jobs by 2031, and a jump in business AI adoption from just over 12% to 60% by 2034.
The domestic strategy matters because it gives Carney credibility on the international stage. Canada has signed 12 international AI partnerships, is building a national public supercomputer and sovereign compute infrastructure, and is expanding the Canadian AI Safety Institute's evaluation capacity. A prime minister pitching a global body while building the domestic institutions to staff it is harder to dismiss than one offering only words.
Why an FSB Clone Might Not Be Enough
Here is the uncomfortable question the proposal must answer: the FSB was built to govern a system that moves in quarters and years. Bank capital ratios do not double in capability every few months. AI does.
The FSB's method — convene, consult, recommend, peer-review, report to the G20 — is deliberate by design. Financial stability work depends on data that arrives with a lag and on consensus that takes months to build. Frontier AI capability, by contrast, is moving fast enough that a coordination body operating on diplomatic time could find itself ratifying norms for a technology that has already moved on. The UN's own timeline illustrates the gap: its High-Level Advisory Body on AI, a 38-expert panel drawn from 68 countries, published its "Governing AI for Humanity" report in September 2024 and found that 118 countries remained excluded from the seven major AI governance initiatives. The first UN Global Dialogue on AI Governance convened in Geneva on July 6-7, 2026 — nearly two years after the report — while model capability continued to compound. A second dialogue is scheduled for May 2027 in New York.
There is also a redundancy problem. The EU AI Office is already doing, in binding form, much of what a new body would propose: documentation requests, model evaluation, corrective orders, fines. The UN's Independent International Scientific Panel on Artificial Intelligence warned in June 2026 that AI could "cause catastrophic harm, either on its own or due to malicious users," while the technology is "outpacing both scientific understanding and governments' ability to adapt." Adding another forum risks creating a talking shop that duplicates existing machinery while the two jurisdictions with real leverage — Washington and Beijing — continue to act bilaterally.
The strongest counter-thesis, then, is that the FSB model is the wrong template for the wrong moment. The real emerging regime may be a two-track system: binding bloc-level regulation in Europe, and national-security-driven export controls and compute chokepoints in the United States and China. In that world, a technology stability board becomes a venue for the countries that do not control the frontier to discuss rules that the countries that do control it have already written elsewhere. The UN General Assembly resolution that established the Global Dialogue on AI Governance — Resolution A/RES/79/325 — pointedly limits the forum to the non-military domain, even as military AI becomes one of the technology's most consequential applications. A body that cannot touch the hardest use cases governs the easy ones.
Carney's answer would be that this is exactly why coordination is needed — that a fragmented regime is the worst outcome for every country that is not a superpower, and that the FSB's post-2008 record shows soft coordination can harden into real compliance when market access is on the line. The test of that answer is not rhetorical. It is empirical.
The Falsifying Signal
The specific signal that would prove the FSB-model thesis wrong is straightforward: if the EU AI Office levies its first major fine against a general-purpose AI model provider within 12 months of its August 2026 enforcement start, and the United States simultaneously expands chip-export controls without joining any new multilateral coordination forum, then the de facto global regime will be bilateral and bloc-level enforcement — not FSB-style consensus. At that point, a technology stability board would be a supplement to the real action, not the centre of it.
Conversely, if the United States signals willingness to participate in a G20-anchored tech coordination body and the EU agrees to channel some AI Office standard-setting through it, Carney's thesis is validated: the middle powers have successfully inserted a multilateral layer into a two-pole system. The 81st session of the UN General Assembly, opening in New York this week, is the first place the idea will be stress-tested; an event at the Carnegie Endowment on September 15 examining "Global Guardrails for Agentic AI" will be another.
What Comes Next: Three Time Horizons
In the short term, the proposal is a positioning move, not a market mover. There is no immediate regulatory shock embedded in a call for a body that does not yet exist. The near-term market question remains the one Amodei raised: whether a slowdown chorus from frontier-lab CEOs weighs on the AI hardware trade. Commentary over the weekend suggested the concern may pressure chipmakers but that the broader AI infrastructure trade remains intact. Investors will watch for whether lab-level caution translates into reduced capital expenditure guidance — that is the transmission channel that matters, not the governance debate.
Over the medium term — the next 12 to 24 months — the shape of the regime will become visible. Watch three things: whether the EU AI Office's first enforcement actions set precedents that other jurisdictions copy; whether the United States joins any multilateral AI coordination forum or doubles down on unilateral export controls; and whether a G20 summit formally endorses a new tech-stability mechanism. The EU's AI Act timeline is itself a clock: full application of the high-risk AI system rules begins August 2, 2027, a year after the GPAI enforcement powers took effect. That date will show whether Europe's model can scale from rulebook to enforcement.
In the long term, the structural question is whether AI governance converges on one of three architectures: a UN-anchored inclusive model (the path the Global Digital Compact and the High-Level Advisory Body point toward), an FSB-style club of major economies (Carney's preference), or a fragmented bloc system (the current drift). The answer determines not only safety outcomes but who captures the economic upside. The global AI market is projected to reach US$4.8 trillion by 2033, according to the Canadian government's own strategy document. Canada's bet — and Carney's personal wager, given his background as a former Bank of England and Bank of Canada governor and a G20 fixture — is that the club model offers the best combination of speed and legitimacy.
The Bottom Line
Carney's proposal is less about AI safety in the abstract than about who writes the rules for the defining technology of the era. An FSB-style technology stability board would let middle powers shape global norms through coordination rather than being forced to choose between American unilateralism and European regulation. It is a sophisticated institutional play, and it fits the moment when the AI industry itself is asking for coordination it cannot organise alone.
But sophistication is not the same as sufficiency. The FSB governed a slow-moving financial system; AI moves at the speed of compute. If the EU's binding enforcement and America's chip leverage become the real regime before a new body proves its weight, Carney's board will be a well-attended forum for rules made elsewhere. The next 12 months of enforcement actions and G20 diplomacy will tell which world we are building — and whether coordination can outrun capability.
Explore more exclusive insights at nextfin.ai.
