We will have done something wrong if we don't take the lesson
Standing in Ireland as the United States moved to restrict access to Anthropic's newest AI models, Canadian Prime Minister Mark Carney offered a warning rooted not in alarm but in hard-won experience: systems that concentrate power in few hands become fragile in ways that only reveal themselves at the worst possible moments. Drawing on his years navigating the 2008 financial collapse, Carney urged governments gathered for the G7 summit to treat this disruption not as a diplomatic grievance but as a lesson in the dangers of technological dependence. The question he left open was whether nations would respond with the sustained, structural effort the moment demands — or simply wait for the next fracture.
- The US export ban on Anthropic's Mythos and Fable AI models sent an immediate signal: governments that built critical infrastructure on a handful of private platforms had handed away control they may not recover easily.
- Carney invoked the 2008 financial crisis not as metaphor but as lived memory — the same logic of interconnected, under-diversified systems that brought down global markets is now threading through AI networks underpinning governments and economies.
- Canada's own vulnerability is acute: years of economic entanglement with the United States, compounded by Trump-era tariffs and Silicon Valley tax disputes, have made the Anthropic restrictions feel less like an isolated incident and more like a pattern.
- At the G7 summit in France, the conversation reached the highest levels — Macron, Amodei, Altman — but Carney was careful to temper expectations, warning that no banner of mission accomplished would emerge from a week of meetings.
- The path forward Carney described is neither retreat from AI nor acceptance of fragility, but a deliberate, long-term strategy of redundancy and diversification — building resilience before the next disruption, not after.
On a Sunday afternoon in Ireland, Mark Carney delivered a warning that felt less like a technology briefing and more like a financial reckoning. The Canadian Prime Minister had just watched the United States impose an export ban on Anthropic's newest AI models — Mythos and Fable — and he saw in that moment something larger than a trade dispute: a system becoming dangerously fragile.
His concern was grounded in experience. As former head of both the Bank of Canada and the Bank of England, Carney had lived through 2008, when tightly interconnected financial networks collapsed in sequence, spreading damage across the globe. The parallel to AI was not perfect, he acknowledged, but it was instructive. A small cluster of firms — Anthropic, OpenAI, and a few others — now underpinned decisions in government, business, and daily life. "We have similar things in terms of model risk," he said.
The timing was deliberate. The G7 summit was convening in France within days, with AI at the center of the agenda. Industry leaders including Dario Amodei and Sam Altman were expected at a Wednesday luncheon. Carney had already spoken with Macron. But he seemed to sense that conversation alone would not be sufficient.
He was careful not to assign blame for the US restrictions. "Nobody's done anything wrong in this situation," he said. His point was not about fault — it was about learning. "We will have done something wrong if we just accept this, don't take the lesson, don't build out and diversify."
The warning fit a pattern that had defined his premiership. Since taking office, Carney had pushed Canada to reduce its economic dependence on the United States — a strategy made urgent by Trump-era tariffs and pressure on Ottawa over Silicon Valley tax policy. The Anthropic ban was the latest expression of how technological leverage and political vulnerability had become inseparable.
Carney was not calling for a slowdown in AI development. What he was calling for was resilience — redundancy and diversity built into the system before the next fracture, not after. He cautioned against expecting the G7 to resolve anything cleanly: "There will not be a mission accomplished banner." What the moment required was sustained effort, the kind that outlasts summits and single policy announcements.
The export ban had exposed something real. Whether governments would treat it as a structural warning or a temporary inconvenience remained, for now, an open question.
Mark Carney stood in Ireland on a Sunday afternoon and delivered a warning that sounded less like a technology forecast and more like a financial reckoning. The Canadian Prime Minister had just watched the United States impose an export ban on Anthropic's newest artificial intelligence models—Mythos and Fable—and he saw in that moment something larger: a system becoming dangerously fragile.
Carney's concern was straightforward. When access to critical technology flows through only a handful of companies, entire economies and governments become vulnerable to disruption they cannot control. He had seen this movie before. As former head of both the Bank of Canada and the Bank of England, he had lived through 2008, when tightly interconnected financial networks collapsed like dominoes, spreading contagion across the globe. The parallel was not perfect, but it was instructive. "We have similar things in terms of model risk," he said, referring to the concentration of artificial intelligence capabilities among a small cluster of firms—Anthropic, OpenAI, and a few others whose systems now underpin decisions in government, business, and daily life.
The timing of his remarks was deliberate. The Group of Seven summit was convening in France within days, and artificial intelligence had moved to the center of the agenda. Carney had already discussed the issue with French President Emmanuel Macron. Industry leaders—including Anthropic's chief executive Dario Amodei and OpenAI's Sam Altman—were expected to attend a G7 luncheon on Wednesday. The conversation was happening at the highest levels, but Carney seemed to sense that the conversation alone would not be enough.
He was careful not to assign blame. "Nobody's done anything wrong in this situation," he said of the US restrictions. Washington had identified what it considered legitimate risks in Anthropic's technology, and the Canadian government acknowledged a "good flow of information" between Ottawa and the American authorities. But Carney's point was not about fault. It was about learning. "We will have done something wrong if we just accept this, don't take the lesson, don't build out and diversify," he said.
This theme had defined his premiership. Since taking office, Carney had pushed Canada toward reducing its economic dependence on the United States—a strategy that had become urgent under President Donald Trump's administration, which had imposed steep tariffs on automobiles, steel, and other sectors while pressing Ottawa to reconsider taxes affecting major Silicon Valley companies. The Anthropic restrictions fit into a larger pattern of technological leverage and vulnerability.
Yet Carney was not calling for a halt to AI development. Countries needed to continue advancing the technology, he said. What they needed instead was resilience built on redundancy and diversity—the opposite of the concentration that had made the system brittle. The G7 summit would be a beginning, not an ending. "There will not be a mission accomplished banner that comes out of the G7," he cautioned. Managing the promises and risks of artificial intelligence would require sustained effort, the kind that could not be resolved in a week of meetings or a single policy announcement.
What Carney was describing was a long-term shift in how governments think about technological power. Not as something to acquire from a few dominant providers, but as something to build, diversify, and distribute. The Anthropic export ban had exposed a vulnerability. The question now was whether countries would treat it as a warning or a temporary inconvenience.
Citações Notáveis
The situation we're in collectively right now with Mythos and Fable is something that can happen with over-reliance on certain models— Mark Carney
We have similar things in terms of model risk. Policymakers should actively seek redundancy and diversity to reduce potential vulnerabilities— Mark Carney
A Conversa do Hearth Outra perspectiva sobre a história
When Carney compares AI concentration to the 2008 financial crisis, what exactly is he worried will break?
The same thing that broke then—a cascade. When everyone relies on the same few systems, and one of them fails or gets cut off, the shock spreads everywhere at once. A bank fails, credit freezes. An AI model gets restricted, and suddenly governments and companies that built their operations around it have no backup.
But the US didn't ban Anthropic entirely. They just restricted access to the newest models. Isn't that a manageable problem?
Only if you have alternatives. If you've already built your infrastructure around Mythos and Fable, a restriction isn't manageable—it's a crisis. And if every country is doing the same thing with the same few models, they're all exposed to the same shock at the same time.
Why does Carney keep saying nobody did anything wrong? That seems like he's softening the criticism.
He's being precise. The US acted within its authority to protect what it saw as national security interests. Anthropic didn't violate any rules. But the system itself—the fact that the world became dependent on a small number of providers—that's the structural problem. You can't blame individuals for a system that was never designed with resilience in mind.
Is he actually expecting the G7 to solve this in a week?
No. He's being realistic about what summits can do. They can set direction, build consensus, maybe commit to funding. But diversifying AI capabilities across multiple countries takes years. It requires investment, talent, infrastructure. A banner doesn't build that.
What does he mean by redundancy and diversity in AI?
Multiple independent systems, developed in different countries, by different companies, using different approaches. So if one gets cut off or fails, others still work. It's the opposite of what we have now—a world where a handful of models do most of the heavy lifting.