US and UK regulators are holding discussions with major banks about cyber security risks linked to Anthropic PBC’s latest AI model, amid wider concern about potential threats to critical financial systems.
In Washington, US Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell convened senior Wall Street executives at the Treasury Department, Bloomberg reported, citing sources.
Go deeper with GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
The meeting was aimed at ensuring large banks understood possible risks associated with Anthropic’s Mythos model – and similar systems that could follow – and that they were taking steps to protect their networks.
Some participants were already in the capital for a Financial Services Forum gathering, a group representing the largest US lenders.
The Treasury did not immediately respond to a request for comment, Bloomberg said. The Federal Reserve declined to comment.
Bloomberg reported that the banks invited are designated as systemically important, meaning disruption at any one of them could have broader consequences for the global financial system.
Some people familiar with the matter said Powell’s presence signalled the issue was being treated as a systemic concern, rather than being tied to the Trump administration’s earlier disputes with Anthropic.
In the UK, the Bank of England, the Financial Conduct Authority and HM Treasury are in urgent contact with the National Cyber Security Centre, and are also engaging with the country’s biggest banks on the issue, the Financial Times reported.
Two people briefed on the discussions told the Financial Times that major UK banks, insurers and exchanges would be warned within the next fortnight about cyber security concerns linked to Anthropic’s latest model, Claude Mythos Preview.
Anthropic has said Mythos is a more advanced model that can locate and exploit weaknesses across major operating systems and web browsers if instructed by a user.
The company has limited initial access to a small group of large technology and financial firms through “Project Glasswing”, which it says is intended to help secure critical systems before similar AI tools are released more widely.
Anthropic has also said it spoke with US officials ahead of the launch about its “offensive and defensive cyber capabilities.”
In one example cited by the company’s security team, Anthropic said it breached a web browser in a way that could allow a malicious website to read information from another site, including “e.g., the victim’s bank.”
Bloomberg reported that those asked to attend the US meeting included Citigroup’s Jane Fraser, Morgan Stanley’s Ted Pick, Bank of America’s Brian Moynihan, Wells Fargo’s Charlie Scharf and Goldman Sachs’ David Solomon.
JPMorgan’s Jamie Dimon did not attend.
