Anthropic’s Mythos AI raises banking cyber risk concerns, FSB briefing planned

Artificial intelligence startup Anthropic is preparing to brief the global financial watchdog on cyber vulnerabilities identified by its latest unreleased AI model, according to a report by the Financial Times.

The report said Anthropic will discuss the capabilities of its new “Mythos Preview” cybersecurity model with members of the Financial Stability Board (FSB), which includes top finance ministries, regulators and central banks from G20 economies. The discussions are being organised following a request from Bank of England Governor Andrew Bailey, who currently chairs the FSB.

Must read: Anthropic Mythos risks force public banks in India to revamp IT investments: Report

The development comes as governments and regulators grow increasingly concerned about how advanced artificial intelligence systems could reshape cybersecurity risks, especially for sectors such as banking that still depend heavily on ageing technology infrastructure.

Anthropic unveiled Mythos earlier this year as a specialised AI system built to detect long-standing vulnerabilities in software, browsers and critical infrastructure. The model has not yet been publicly released and is currently being tested under controlled access arrangements.

Cybersecurity researchers and policymakers have warned that such highly capable systems could become a double-edged sword helping organisations identify weaknesses faster, while also potentially enabling sophisticated cyberattacks if misused.

Bailey had publicly flagged concerns around the model during an event at Columbia University in New York last month, shortly after Anthropic announced the model.

Must read: BT explainer: Why Anthropic’s Claude Mythos could become a major security risk?

“It would be reasonable to think that the events in the Gulf are the most recent challenge to us in this world, until, I think it was last Friday, you wake up to find that Anthropic may have found a way to crack the whole cyber risk world open,” Bailey said at the event.

“The issue is: to what extent is this new version of the product going to be able to, in a sense, identify vulnerabilities in other systems which can be exploited for cyber attack purposes,” he added.

The planned FSB briefing highlights how frontier AI systems are increasingly becoming a topic of concern not just for technology regulators, but also for financial stability authorities worried about systemic cyber risks.

Anthropic has been positioning Mythos as a defensive cybersecurity tool designed for vetted organisations and government partners. Earlier this month, the Pentagon also confirmed that it was deploying the model under a restricted programme aimed at identifying and patching software vulnerabilities across US government systems.