world-economy

Anthropic Withholds Powerful AI Model Over Banking Security Risks

Anthropic has decided against releasing Claude Mythos, a newly developed AI model, citing security concerns for financial institutions. The company's cybersecurity chief Michael Moore said the model demonstrated advanced capabilities in chaining together multiple software vulnerabilities into functional exploits. Unlike previous Claude iterations, Mythos could find disparate code flaws that individually pose minimal risk but collectively create significant exposure when combined. The discovery prompted Anthropic to develop a security tool addressing software vulnerabilities rather than deploy the model publicly. Banking officials have raised concerns about the potential implications. The decision underscores growing caution among AI developers regarding the release of increasingly sophisticated systems.