Bank of England Governor Says UK Banks Blocked From Anthropic's Mythos AI Tool

Bank of England Governor Andrew Bailey stated that British banks remain unable to access Anthropic PBC's powerful AI tool Mythos, according to Bloomberg reports of his remarks. The restriction leaves UK financial institutions at a disadvantage as peers elsewhere potentially deploy the advanced system for cybersecurity and risk management.

Bailey's comments came during a Bloomberg interview with Stephanie Flanders at a central banking event on May 29, 2026. The governor did not specify whether the block stems from regulatory barriers, commercial licensing disputes, or national security reviews. His candor about the gap stands out, as central bankers rarely discuss specific vendor access limitations so directly in public forums.

The remarks signal growing tension between national financial regulators and the concentrated control of frontier AI systems by a handful of American firms.

Rather than resolving the access issue unilaterally, Bailey called for a multinational approach to fighting cybersecurity threats. This framing treats AI tool access as a collective governance challenge rather than a procurement problem for individual institutions.

His stance aligns with broader themes from his recent speeches. In a February 2025 address at the University of Chicago Booth School of Business, Bailey stressed that financial stability assessments had not fully absorbed the scale of market structure changes. He highlighted the Bank of England's System Wide Exploratory Scenario as a pioneering effort to map risks across non-bank financial entities. The throughline is consistent: Bailey views modern financial threats as systemic and cross-border, requiring coordinated institutional responses rather than isolated national fixes.

The multinational appeal also reflects practical constraints. No single regulator can compel a private AI developer like Anthropic to grant access to foreign banks.

Bailey has increasingly framed artificial intelligence as integral to regulatory capacity, not merely a sectoral concern. Reuters reported his view that AI can help regulators perform their oversight functions more effectively, though details of that application remain unspecified.

This represents an evolution in central bank thinking. Traditional supervision relies on periodic reporting and static stress tests. AI-enabled monitoring promises real-time pattern detection across sprawling, interconnected markets. The Bank of England's own innovations, including the System Wide Exploratory Scenario, suggest an appetite for more dynamic analytical tools.

Yet the Mythos access gap exposes a paradox: regulators want AI to strengthen oversight while their domestic banks cannot obtain the most capable systems. Bailey's public complaint implies frustration with this asymmetry and perhaps pressure from UK institutions who see foreign competitors gaining operational advantages.

The specific naming of Anthropic's Mythos underscores how concentrated AI influence has become in financial infrastructure. Anthropic PBC, founded by former OpenAI researchers and backed by Google and Amazon, has positioned its systems as enterprise-grade alternatives to consumer-facing chatbots. Mythos appears to represent a specialized or more capable tier of their offerings, though technical specifications have not been publicly detailed.

For UK banks, exclusion from Mythos could affect fraud detection, transaction monitoring, and cybersecurity operations where Anthropic's models reportedly outperform general-purpose alternatives. The competitive implications extend beyond direct productivity: regulators increasingly expect institutions to deploy best-available risk management technology, and gaps may attract supervisory scrutiny.

The episode also highlights how AI vendors now wield geopolitical leverage through access decisions. Anthropic's choices about which markets and institutions to serve carry systemic consequences.

Bailey's public intervention likely precedes formal regulatory or diplomatic action. The Bank of England could pursue several paths: negotiating a UK-specific access agreement with Anthropic, pressing for EU or G7 coordination on AI vendor access standards, or accelerating domestic alternatives through partnerships with British AI firms like DeepMind or smaller startups.

The multinational framework Bailey advocated would face significant implementation hurdles. Divergent national AI regulations, data sovereignty requirements, and competing industrial interests complicate collective bargaining with American technology companies. The UK post-Brexit position adds complexity, no longer able to rely on EU bloc negotiating power.

Market participants should expect continued volatility in AI access arrangements. Bailey's remarks suggest the Bank of England will treat this as a priority issue, but resolution timelines remain uncertain given the commercial and political variables involved.