UK banks remain unable to access Anthropic's Mythos AI model for cybersecurity testing, exposing growing tensions between AI governance, national security and global financial stability.
 |
| The delay in granting UK banks access to Anthropic's AI model highlights how advanced artificial intelligence is becoming intertwined with geopolitics, regulation and critical infrastructure protection. Image: FC |
FC Desk — May 30, 2026:
The debate surrounding artificial intelligence is rapidly expanding beyond innovation and productivity. Increasingly, it is becoming a matter of national security, financial stability and international politics.
That reality was underscored this week when Bank of England Governor Andrew Bailey revealed that British banks still have not gained access to Anthropic's Mythos AI model, despite concerns raised more than six weeks ago about its potential implications for cybersecurity.
The delay is noteworthy because Anthropic itself appears willing to provide trial access to the model. According to Bailey, the obstacle is not technical but political, with the approval process seemingly caught up in discussions involving the U.S. administration.
The situation illustrates a growing challenge facing governments and regulators worldwide. As AI systems become more powerful, access to them is increasingly treated as a strategic issue rather than a purely commercial one.
For banks, the concern is straightforward.
Financial institutions face an escalating wave of cyber threats, ranging from ransomware attacks to sophisticated attempts to breach critical infrastructure. Regulators want banks to understand both the risks and capabilities of advanced AI systems before those technologies become widely available.
Testing systems against cutting-edge AI models is increasingly viewed as part of modern cybersecurity preparedness. Without access, regulators may struggle to assess potential vulnerabilities or prepare institutions for future threats.
Bailey's concerns carry particular weight because he also chairs the Financial Stability Board, the international body responsible for monitoring risks to the global financial system.
His warning reflects a broader fear that AI-driven cyber threats could create systemic risks extending beyond individual banks or national borders.
Modern banking systems are deeply interconnected. A significant cyberattack on one institution can quickly affect payment networks, financial markets and cross-border transactions. In such an environment, weaknesses in one jurisdiction can become problems for many others.
That is why Bailey argues that cyber risk cannot be managed through isolated national strategies.
His comments point to a larger policy dilemma. Governments want to encourage AI innovation while also preventing advanced models from being misused. However, restrictions designed to limit risks may also prevent trusted organizations from accessing tools that could strengthen defenses.
The controversy surrounding Mythos AI highlights how difficult that balance has become.
Anthropic has already been involved in discussions with U.S. policymakers over guardrails governing the use of advanced AI technologies, particularly in military and national security contexts. These debates have intensified as governments increasingly view frontier AI models as assets with strategic importance.
Adding another layer of complexity, some cybersecurity experts have challenged earlier warnings about the model's capabilities. While concerns initially emerged that Mythos could dramatically expand hacking capabilities, several researchers now argue that fears of unrestricted AI-powered cyberattacks may be overstated.
This disagreement reflects a wider divide within the cybersecurity community.
Some experts believe advanced AI models could significantly lower barriers for malicious actors, enabling faster and more sophisticated attacks. Others argue that existing technical limitations and security controls reduce those risks, at least for now.
Regardless of where the truth ultimately lies, policymakers appear unwilling to take chances.
The timing is also significant because President Donald Trump recently postponed an executive order that was expected to establish a voluntary framework for AI developers to engage with the U.S. government before releasing advanced models to the public.
That delay has added uncertainty to an already complex regulatory landscape.
The result is a situation where financial regulators are seeking access to powerful AI systems for defensive purposes, technology companies are navigating evolving government requirements, and policymakers are still determining how advanced AI should be governed.
The Mythos case may therefore represent more than a dispute over access to a single model.
It is an early example of how artificial intelligence is becoming intertwined with global power, economic security and critical infrastructure protection. The question is no longer just what advanced AI can do. Increasingly, it is who gets access to it, when they get access, and under whose authority.
For regulators such as Bailey, the answer matters because cyber threats are global. If advanced AI can influence the security of financial systems, then access to that technology may become just as important as the regulations designed to control it.
As governments, technology companies and financial institutions continue to navigate this new reality, the battle over Mythos AI could offer a glimpse into the future of international AI governance—one where cybersecurity, geopolitics and financial stability are becoming impossible to separate.