AI is transforming from a mere support mechanism into an essential tool for banks to appreciate and secure against. In the United States, several major financial institutions are examining a sophisticated AI model that possesses capabilities resembling those of cyber attackers. This initiative, involving calls for evaluations of such technology, has not solely originated from banks.

Officials connected to Donald Trump’s administration are advocating for financial institutions to rigorously assess how this AI model could unveil vulnerabilities within their systems before adversaries employ it against them. This concern is particularly centred around Mythos, an advanced AI model created by Anthropic.

Capabilities of the Mythos AI Model

Mythos distinguishes itself from traditional cybersecurity tools by mimicking the thought processes of an attacker. This innovative approach enables it to scan banking systems for hidden vulnerabilities and devise potential methods for exploiting these weaknesses. Such dual capabilities—both defensive and offensive—have piqued the interest of regulators concerned about the security of financial institutions.

So far, access to Mythos has been restricted, with a few institutions like JPMorgan Chase being among the pioneers in this exploration. Other major banks, including Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley, are reportedly in the initial stages of internal testing.

The urgency surrounding this initiative was highlighted in a recent meeting in Washington, where US Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell summoned senior executives from Wall Street firms. Instead of identifying a specific looming threat, officials cautioned that banks need to begin stress-testing their infrastructures against the advanced capabilities of AI before they become readily available.

Growing Concerns Over Cybersecurity Risks

The shift in approach underscores an escalating concern among policymakers regarding cybersecurity. While risks in the banking sector have long been recognised, AI has the potential to accelerate and amplify the frequency of cyber attacks. Unlike previous tools, Mythos is capable of identifying not just isolated vulnerabilities but also interconnecting minor weaknesses into substantial threats.

This “chained vulnerability” strategy has presented challenges even for adept human hackers, making it imperative for banks to adapt. Anthropic has conducted internal evaluations that reveal some of these potential risks; for instance, the AI has successfully identified means to exploit web browsers to access sensitive data, including financial details.

The limited rollout of Mythos reflects the need for caution. Only a select group of organisations have received early access through a programme designated “Project Glasswing.” In addition to banks, tech giants like Amazon and Apple have also been involved, with the objective of fortifying critical systems and comprehending the AI model’s behaviour prior to broader release.

Officials underscored the urgency of these evaluations. Kevin Hassett commented on the appropriateness of Secretary Bessent’s actions, reinforcing the commitment to ensuring security against potential threats. He stated that measures are being taken to guarantee public safety, including the decision by Anthropic to postpone the public release of the model until government assessments are complete.

Long or Short, get news the way you like. No ads. No redirections. Download Newspin and Stay Alert, The CSR Journal Mobile app, for fast, crisp, clean updates!

App Store –  https://apps.apple.com/in/app/newspin/id6746449540 

Google Play Store – https://play.google.com/store/apps/details?id=com.inventifweb.newspin&pcampaignid=web_share