War and an historic energy shock are tough financial stability risks to compete with.
Yet Anthropic’s as-yet unreleased AI model Mythos added a layer of dread to conversations among policymakers gathered in Washington at two institutions created eight decades ago to help foster global economic harmony.
At the spring meetings of the International Monetary Fund and the World Bank, the military conflict in the Middle East dominated the talks and forced governments to dust off crisis response plans, pledge cooperation where possible, and bolster consumption with fiscal and monetary tools. With Mythos, there were far more questions than answers.
"If it falls in the wrong hands, it could be really bad,” European Central Bank President Christine Lagarde told Bloomberg Television on Tuesday. She was referring to Mythos’ destructive potential, echoing other central bankers, finance ministers, regulators and investment chiefs hunting for information on potential threats and safeguards.
The broad outlines of the issue were thrust into the public domain by a Bloomberg News report revealing that Treasury Secretary Scott Bessent had gathered Wall Street leaders to warn on an imminent AI model that could herald a new era of cyber attacks launched autonomously by bots.
"I feel confident that everyone is now on board, rowing in the same direction to build up resiliency,” Bessent told CNBC on Wednesday.
Outside the US, though, the seeming urgency of that meeting helped fuel a state of high alert among senior officials and bankers. Some of them fretted over a technology that could breach traditional cyber defences, leaving the financial system open to untold threats.
"It would be reasonable to think that the events in the Gulf are the most recent challenge to us in this world,” Bank of England Governor Andrew Bailey said at an event at Columbia University in New York earlier this week. That is, until "you wake up to find that Anthropic may have found a way to crack the whole cyber risk world open and you think, what did I do wrong in a past life.”
Main Questions
The challenge facing Bailey and other leading financial stability experts is that they don’t know enough about the threats to understand how big an advance Mythos really is on the escalating cybersecurity risks they have been warning about for years now.
Could Mythos lead to the mass looting of bank accounts, paralysis in international payment systems or spark a crisis of confidence that shatters the bedrock of the financial system – trust?
Is this akin to the risks from quantum computing, a known threat which regulators are already studying in detail, or will it require a new playbook?
Such questions circulated through the corridors and conference rooms of Washington’s Willard Hotel, where Abraham Lincoln lived before his inauguration and where the Institute of International Finance this week hosted senior officials and executives.
Many remain unanswered.
The extent to which San Francisco-based Anthropic is sharing details of and access to Mythos with non-US officials and banks wasn’t immediately clear. One American leader of a European bank said they’d been promised a briefing. The limited access may leave some of the world’s financial system incapable of following the International Monetary Fund’s advice to stay "at the frontier” of developments in cyber threats.
Canada’s Finance Minister Francois-Philippe Champagne said Mythos "requires our full attention” and that he wants to raise the issue with his counterparts. "We have a common interest to ensure the resiliency of our financial system,” he said in an interview Wednesday.
The US Treasury Department’s technology team is seeking to gain access to Anthropic PBC’s Mythos AI model so it can begin hunting for vulnerabilities, a person familiar with the situation told Bloomberg this week.
John Williams, president of the Federal Reserve Bank of New York, said on Thursday that recent advances "have demonstrated the power of these AI tools, both at identifying vulnerabilities or potentially exploiting these vulnerabilities, has moved a lot faster than many expected.” The New York Fed is a key nerve centre for global financial transactions and oversight of stability risks.
FSB’s Role
Financial stability threats are traditionally coordinated through the Financial Stability Board, currently chaired by the BOE’s Bailey. But the FSB has mostly taken a back seat on cyber risks, because some members are reluctant to be transparent in a broad group that includes countries like China. That leaves cyber largely in the hands of narrower groups like the Group of Seven industrial economies.
At a G-7 finance chiefs meeting Wednesday, the need for an international institutional framework to oversee the governance of AI was discussed, on the basis that no country could address the Mythos risks alone, according to two people familiar with those talks.
There was support for exploring the topic, given the exponential growth of the risks and their potentially severe consequences, one of the people added, though the route forward is unclear.
Adding to the worries is the heightened geopolitical risks that accompany the doomsday AI scenario, where the technology flips from transforming the financial sector’s efficiency to obliterating its technology infrastructure. The Middle East conflict increases the perceived threat of state-sponsored hacks.
‘Real Urgency’
Officials and bankers alike are also mindful that if such destructive AI technology can be created by a US company, it could likely be created by one in a more hostile nation, which would not give Wall Street a heads up on the looming threat.
"Frontier AI capabilities are advancing faster than governance frameworks are being designed to control them,” said Laila Khawaja, research director of Gavekal Technologies, who covers US-China tech. "There is real urgency to build guardrails before these models become deeply embedded in critical systems like finance.”
This week’s quest for answers also highlights the gulf between AI’s haves and have-nots, and the splintering of the global financial stability ecosystem. Europe is increasingly concerned about its over-reliance on foreign tech companies.
Officials also privately lament the "every man for himself” approach that is increasingly creeping into global conversations on finance, and could hinder the effective sharing of information on risks globally. Several European officials said they would use this week’s meetings to encourage their US peers to share what they can.
Elisabeth Svantesson, Sweden’s finance minister, told Bloomberg News that there’s an "an early-warning meeting” later Thursday with central bankers and ministers, and AI will be a topic.
"My message will be that AI is fantastic but unpredictable leadership together with AI is dangerous,” she said, adding that the US and Israel’s war and uncertainty are the predominant themes, along with cyber threats. – Bloomberg
