Finance ministers, central bankers, and leading financial experts have raised serious concerns about a powerful new AI model that could threaten global financial systems.
The model, called Claude Mythos, was developed by Anthropic. Early testing shows it can detect and exploit weaknesses in major software systems. This has triggered urgent discussions among policymakers and financial leaders.
Global Leaders Take the Threat Seriously
At a recent meeting hosted by the International Monetary Fund in Washington DC, officials discussed the risks linked to this AI system.
Canada’s Finance Minister, François-Philippe Champagne, said the issue demands global attention. He explained that unlike physical threats, AI risks are harder to predict and control.
He stressed the need for strong safeguards. Governments want to ensure financial systems remain stable and secure against emerging digital threats.
What Makes Claude Mythos Different
Claude Mythos is part of Anthropic’s Claude AI family. It competes with systems like ChatGPT and Gemini.
Researchers testing the model found it highly skilled in cyber-security tasks. It can uncover hidden software bugs and suggest ways to exploit them.
Due to these risks, Anthropic has not released Mythos to the public. Instead, it shared limited access with major tech firms such as Microsoft, Nvidia, Amazon Web Services, and CrowdStrike.
This initiative, called Project Glasswing, aims to strengthen global software security before wider deployment.
Experts Urge Caution, Not Panic
Some cyber-security experts believe the concerns may be slightly overstated. They point out that the model has not undergone broad independent testing.
The UK AI Security Institute reviewed a preview version. Its report confirmed that Mythos performs well in weak security environments. However, it does not appear vastly more advanced than earlier models like Claude Opus.
Researchers also noted that similar AI tools will likely emerge in the near future. This suggests the issue is not limited to a single model.
Banks Prepare for Real-World Testing
Financial institutions are now preparing to test their systems against this AI.
C. S. Venkatakrishnan, CEO of Barclays, said the situation requires close attention. He emphasized the need to identify vulnerabilities quickly and fix them before they can be exploited.
