WASHINGTON—It isn’t just the U.S. government and U.S. financial institutions raising their concerns about the risks posed by advanced artificial intelligence models that can expose vulnerabilities in financial institutions’ cyber defenses. Banks and bank regulators in other countries are just as worried.
At last week’s International Monetary Fund and World Bank spring meetings, central bankers and regulators focused on a new model developed by Anthropic known as Claude Mythos Preview and its potential to identify weaknesses in banking systems at unprecedented speed and scale.
The CU Daily reported on an emergency meeting between the U.S. government and the nation’s biggest banks here.
Andrew Bailey, governor of the Bank of England and chair of the Financial Stability Board, described the issue as “a very serious challenge for all of us,” the Financial Times reported, adding that regulators must move quickly to assess the threat.
Officials are increasingly concerned that large language models could chain together software vulnerabilities faster than human analysts, potentially shifting the advantage toward cyber attackers.
Thousands of Vulnerabilities
Anthropic said earlier this month that its Mythos model had identified thousands of high-severity vulnerabilities, including flaws in major operating systems and web browsers, according to the Financial Times. Access to the model has been limited to about 40 companies, including Amazon, Apple and JPMorgan Chase, to allow testing and remediation.
The Financial Times reported that the issue is evolving beyond a technical challenge into a broader governance concern, as policymakers grapple with how to oversee rapidly advancing AI capabilities.
‘Lack of Coordinated Oversight’
Christine Lagarde, president of the European Central Bank, told Bloomberg TV there is currently no framework “to actually mind those things,” underscoring the lack of coordinated oversight.
Regulators face a balancing act, seeking to avoid slowing a technology with significant economic potential while also trying to prevent systemic risks from emerging unchecked. The Financial Times noted that geopolitical tensions and uneven access to advanced models could complicate efforts to coordinate a global response.
Additional reporting cited by the Financial Times from PYMNTS indicated that the U.S. Treasury Department is seeking access to Anthropic’s Mythos model, while U.K. regulators are also examining potential threats to financial stability. At the same time, Anthropic is experiencing rapid commercial growth, with enterprise demand accelerating.
‘Split-Screen Reality’
The developments highlight what the Financial Times described as a “split-screen” reality: AI firms are gaining traction in financial services even as regulators and institutions race to understand and manage the risks posed by increasingly powerful tools.
