UK banks remain unable to access Anthropic’s Mythos artificial intelligence system for testing against cyber risks, six weeks after initial concerns emerged, according to Bank of England Governor Andrew Bailey. Speaking to Bloomberg TV on Friday, Bailey indicated that Anthropic had offered the model for limited trials, yet a political delay appeared to be involved. He noted the matter seemed tied to processes with the U.S. administration during an interview at a central banking event in Reykjavik. Bailey added that the varying timelines across companies were unclear, but access was essential given the risks. The company has previously clashed with U.S. officials over rules for military applications of its AI tools. Bailey had previously highlighted the model’s potential to transform cyber risk assessment. Some experts have since suggested that concerns about unrestricted misuse may be exaggerated. Last week, President Donald Trump delayed an executive order on AI that would have introduced a voluntary framework for developers to consult the government ahead of releasing advanced models. Bailey, who also leads the Financial Stability Board, called for coordinated international efforts on cyber threats, stressing that national measures alone would fall short due to global interconnections in banking.
