Mark Carney warns against AI dependence after US blocks access to Anthropic’s latest models

Canada’s Prime Minister Mark Carney has warned that the US decision to restrict overseas access to Anthropic PBC’s newest artificial intelligence models should serve as a wake-up call for governments and businesses increasingly dependent on a handful of advanced AI systems.

Speaking to reporters during a visit to Ireland on Sunday, Carney said the disruption highlighted the dangers of concentrating too much power in a few technologies, according to Bloomberg. His remarks come as artificial intelligence emerges as a major theme ahead of this week’s Group of Seven (G7) leaders’ summit in France, where political leaders and industry executives are expected to discuss the technology’s opportunities and risks.

A lesson in avoiding concentration

Referring to Anthropic’s latest models, Mythos and Fable, Carney said the episode demonstrated how vulnerable countries could become when access to critical technologies rests with only a small number of providers.

“The situation we’re in collectively right now with Mythos and Fable is something that can happen with over-reliance on certain models,” he said.

More from Tech

Carney stressed that no single party was to blame for the current situation, but argued that governments would be making a mistake if they failed to learn from it.

“Nobody’s done anything wrong in this situation, but we will have done something wrong if we just accept this, don’t take the lesson, don’t build out and diversify,” he said.

The comments reflect a broader theme that has defined Carney’s premiership. Since taking office, the Canadian leader has repeatedly championed diversification, particularly in response to Canada’s heavy economic dependence on the United States.

That strategy has gained urgency under President Donald Trump’s administration, which has imposed steep tariffs on several sectors, including automobiles and steel. Washington has also pressed Ottawa to reconsider taxes affecting major Silicon Valley companies.

Despite the latest restrictions involving Anthropic’s AI models, Carney said communication between Ottawa and Washington remained constructive. He noted that there was a “good flow of information” between both governments regarding AI and acknowledged that US authorities had identified “some risks” associated with Anthropic’s latest technology.

Echoes of the financial crisis

Carney, who previously headed both the Bank of Canada and the Bank of England and spent years at Goldman Sachs Group, likened the challenges posed by AI concentration to the interconnected vulnerabilities exposed during the 2008 global financial crisis.

The collapse of major financial institutions at the time revealed how systemic risks could spread rapidly through tightly linked networks. Similar concerns, he suggested, now apply to artificial intelligence.

“We have similar things in terms of model risk,” Carney said, adding that policymakers should actively seek “redundancy and diversity” to reduce potential vulnerabilities.

His comments underline growing international concerns about the concentration of AI capabilities among a small number of firms, including Anthropic and OpenAI, whose technologies are increasingly becoming central to governments, businesses and consumers.

Artificial intelligence is expected to occupy a prominent place on the agenda when G7 leaders gather in France in the coming days. Carney said he had already discussed the issue with French President Emmanuel Macron.

Several leading figures from the AI industry are also expected to participate in a G7 luncheon on Wednesday, including Anthropic chief executive Dario Amodei and OpenAI chief executive Sam Altman, Bloomberg reported.

While Carney said countries must continue advancing the technology, he cautioned against expecting immediate breakthroughs from the summit.

“We need to make progress,” he said, while warning that “there will not be a mission accomplished banner that comes out of the G7.”

Instead, the Canadian leader suggested that managing the promises and risks of artificial intelligence would require a sustained effort, one built not on dependence, but on resilience and diversity.