U.S. Treasury and Federal Reserve Convene Critical Meeting on AI Cyber Threats
(April 11)In a decisive move, U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell convened top executives from Wall Street to discuss mounting apprehensions regarding Anthropic PBC’s latest artificial intelligence (AI) model. The new system is perceived as a catalyst for an unprecedented surge in cyber vulnerabilities.
Bessent and Powell organized the session at the Treasury’s Washington headquarters on Tuesday, aiming to enlighten banks about the prospective threats introduced by Anthropic’s Mythos and analogous models.
Sources privy to the discussions, who chose to remain anonymous due to the confidential nature of the meeting, indicated that the objective was to ensure banks undertake the necessary measures to fortify their cybersecurity infrastructures.
A number of executives were already in the vicinity for the Financial Services Forum gathering, an advocacy collective representing the largest financial institutions. Requests for comments from Treasury representatives went unanswered, while a Fed spokesperson declined to provide any insights.
The previously unreported meeting, arranged with little advance notice, underscores regulatory concerns about a new wave of cyberattacks as a paramount risk to the financial sector.
All banks present are classified as systemically important by leading regulatory bodies, highlighting their pivotal role in maintaining global financial stability.
Powell’s involvement signaled that the urgency of the situation pertained to systemic risks rather than prior tensions between the Trump administration and Anthropic, as noted by some attendees.
The Fed’s extensive network of examiners is well-acquainted with banking operations, enhancing the gravity of these discussions.
National Economic Council Director Kevin Hassett remarked on Fox News last Friday, We are taking every measure possible to ensure that everyone is shielded from these burgeoning risks.
Anthropic has agreed to delay the public rollout of the model until officials can thoroughly assess the implications. There exists an undeniable sense of urgency.
Other regulators are equally proactive. Reports indicate that the Bank of Canada, along with the nation’s prominent banks and financial bodies, convened last Friday to deliberate on cybersecurity threats posed by Mythos.
The Bank of England is also planning a meeting with senior bank and insurance executives to discuss their preparedness.
Anthropic’s Mythos emerges as a formidable AI system capable of discerning and potentially exploiting weaknesses across all major operating systems and web browsers when commanded by a user.
This regulatory caution aligns with Anthropic’s own approach. The company has restricted the initial dissemination of Mythos to a select group of leading technology and financial firms, including Amazon.com Inc., Apple Inc., and JPMorgan Chase & Co.
These entities are participating in “Project Glasswing,” which aims to secure critical systems ahead of the widespread availability of similar AI models.
Anthropic has confirmed discussions with U.S. officials regarding Mythos and its dual capabilities in both offensive and defensive cyber operations prior to its recent limited release.
In its controlled rollout, Anthropic highlighted several exploitative vulnerabilities its AI system can identify during testing.
Although none pertained specifically to the financial sector, the firm’s security team noted an incident where they compromised a web browser, facilitating a hacker’s website to extract data from another site, specifically that of a “victim’s bank.”
Mythos Preview allegedly “autonomously discovered” methods for extracting information from “multiple different web browsers,” subsequently utilizing this capability to identify vulnerabilities, according to communications from Anthropic’s security team.
Furthermore, Anthropic is currently engaged in a legal struggle against the Trump administration, which has categorized the company as a supply-chain risk—an assertion Anthropic vehemently contests.
A recent ruling from a federal appeals court postponed, at least momentarily, Anthropic’s request to halt the Pentagon’s classification.
Among the CEOs summoned for discussions with the Fed and Treasury were Citigroup Inc.’s Jane Fraser, Morgan Stanley’s Ted Pick, Bank of America Corp.’s Brian Moynihan, Wells Fargo & Co.’s Charlie Scharf, and Goldman Sachs Group Inc.’s David Solomon. Notably, JPMorgan’s Jamie Dimon was unable to attend.
Bank representatives, including those from Anthropic, refrained from commenting following the meeting.
In recent years, regulatory bodies have mandated that banks maintain capital reserves commensurate with potential cyberattack risks, alongside other so-called operational risks, such as lawsuits and employee misconduct.
Banks have occasionally expressed resistance to these requirements, citing the inherent difficulty in quantifying operational risk compared to the more tangible market and credit risks that influence their capital adequacy.
Source link: Theedgemalaysia.com.
