Anthropic's 'Mythos' AI Gains UK Bank Access Amid Global Regulatory Scramble

A new AI model considered too hazardous for public release is being made available to UK banks. Anthropic, the creator of Claude AI, will grant financial institutions access to its 'Mythos' model within days, expanding a program previously limited to major US tech firms.

The company itself has warned that Mythos represents an unprecedented risk due to its sophisticated ability to discover and exploit software weaknesses, potentially surpassing elite human coders. This capability, while powerful for defense, raises severe concerns for economic and national security if misused.

The timing coincides with heightened anxiety among global financial leaders. At the IMF and World Bank meetings in Washington, officials grappled with how to approach such advanced AI. Canadian Finance Minister François-Philippe Champagne framed it as an 'unknown unknown,' contrasting it with more tangible geopolitical risks. Bank of England Governor Andrew Bailey acknowledged the speed of AI development presents a 'very serious challenge,' while questioning the optimal moment to establish rules—too early may stifle innovation, too late may lose control.

European Central Bank President Christine Lagarde pointed to the dual-use dilemma: a tool from a 'responsible company' that could be 'really good' or 'really bad.' She noted a lack of existing governance frameworks to manage these specific threats.

With the US Treasury already convening meetings with major bank chiefs on the issue, UK regulators are set to follow suit. As the International Monetary Fund's Dan Katz stated, the cybersecurity risks posed by such digital evolution will be essential to the international agenda in the coming months.

Source: The Guardian