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

Anthropic, an innovative AI startup, is set to inform the Financial Stability Board (FSB) about critical cyber vulnerabilities linked to its forthcoming AI model, as reported by the Financial Times. The discussion will focus on the new "Mythos Preview" cybersecurity model, with participation from key financial leaders, regulators, and central bank officials from the G20 nations. The initiative follows a request from Andrew Bailey, Governor of the Bank of England and current chair of the FSB. As concerns mount regarding how advanced AI systems could alter the landscape of cybersecurity risks—particularly within the banking sector, which still relies on outdated technological frameworks—this briefing has become increasingly pertinent. Earlier this year, Anthropic introduced Mythos, a specialized AI designed to uncover long-standing vulnerabilities in software, browsers, and critical infrastructures. While the model has not yet been publicly launched, it is undergoing testing under controlled conditions. Experts in cybersecurity and policymakers have raised alarms that such powerful systems could serve as a double-edged sword—enabling organizations to swiftly identify weaknesses while also potentially facilitating advanced cyberattacks if mismanaged. During a recent event at Columbia University, Bailey expressed his concerns about the implications of Mythos, suggesting that it may radically change the dynamics of cyber 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," he commented. He emphasized the critical question: to what extent could this new product identify vulnerabilities that might be exploited for malicious purposes? The forthcoming FSB briefing underscores the urgency with which both technology and financial stability authorities are addressing the implications of cutting-edge AI systems. Anthropic has been marketing Mythos as a defensive tool aimed at trusted organizations and government partners. Notably, just this month, the Pentagon confirmed its plans to implement the model in a restricted capacity to identify and rectify software vulnerabilities within US government systems.