Bessent, Powell warned bank CEOs about Anthropic model risks: Reuters

Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell convened a meeting with major U.S. bank chief executives in Washington, D.C., this week to issue a formal warning about the potential cybersecurity and systemic risks posed by the advanced artificial intelligence models developed by the company Anthropic. The meeting, which included leaders from institutions such as JPMorgan Chase, Bank of America, and Citigroup, was prompted by growing concerns within the federal government that the rapid, widespread adoption of these powerful AI systems within the financial sector could introduce unprecedented vulnerabilities. The high-level discussion focused on the specific dangers associated with Anthropic's Claude models and similar frontier AI. Officials are particularly worried about sophisticated AI being used to orchestrate complex cyberattacks, generate highly convincing phishing and social engineering campaigns, or exploit vulnerabilities in financial networks at a scale and speed beyond human capability. There is also apprehension about AI models being manipulated to produce flawed financial analyses or automated trading decisions that could destabilize markets. The meeting served as a direct channel for regulators to stress that banks must treat AI integration not just as a technological upgrade but as a core component of their enterprise risk management frameworks. This regulatory engagement marks a significant escalation in the government's proactive oversight of AI in critical infrastructure. It reflects a shift from observing technological trends to actively managing the associated national security and economic stability threats. The warning to bank CEOs indicates that federal authorities are moving to establish guardrails before potential crises emerge, urging the financial industry to conduct rigorous internal audits of their AI systems, enhance defensive capabilities, and ensure human oversight remains integral to high-stakes processes. The action underscores the Biden administration's broader strategy to mitigate risks from advanced AI while balancing its potential for innovation.