Bank of England Governor Andrew Bailey said UK banks still do not have access to Anthropic's Mythos AI model [1].

The restriction highlights a growing tension between the rapid deployment of generative AI and the need for stringent cybersecurity safeguards in the financial sector. As banks seek to integrate advanced models to improve efficiency, regulators fear that unvetted AI could introduce systemic vulnerabilities to the UK economy.

Bailey said Tuesday, April 14, 2026, regarding the risks associated with the technology [1]. He said that AI models like Mythos pose growing cybersecurity risks to financial institutions [1]. Because these risks can cross borders, Bailey called for an international regulatory approach to ensure stability [1].

While UK institutions remain without access, other regions may be closer to integration. Three people familiar with Anthropic's plans said access to the Mythos model would be offered to European banks soon [2], [3].

There are conflicting reports regarding the immediate timeline for the UK. While some reports indicate banks still lack access [1], other sources suggest UK banks will receive a briefing on the Mythos model within days [4]. This discrepancy underscores the cautious pace at which the Bank of England is permitting the adoption of high-capacity AI tools within the domestic banking infrastructure.

Bailey's focus remains on the potential for AI to be weaponized or to create unforeseen gaps in bank defenses. By urging regulators to assess these risks, the Bank of England is signaling that security audits will likely precede any widespread rollout of the Mythos model in the UK financial market [1].

“UK banks still do not have access to Anthropic's Mythos AI model”

The Bank of England is prioritizing systemic stability over rapid technological adoption. By restricting access to the Mythos model until cybersecurity risks are quantified, the UK is positioning itself as a cautious regulator in the global AI race, potentially lagging behind European peers in deployment but aiming to avoid a catastrophic AI-driven cyber event.