July 14, 2026, (Inside AI) — Britain's largest banks remain locked out of Anthropic's advanced AI model Mythos, a tool critical for cybersecurity defense, exposing a dangerous dependency on U.S. technology. Harriet Rees, the government's appointed AI champion and CIO of Starling Bank, warns the situation demands immediate action to build sovereign AI capabilities.
The exclusion follows Anthropic's April release of Mythos to a handpicked group, including JPMorgan in the U.S. UK banks crave the model for its unmatched ability to spot and patch security flaws, but only a handful of American lenders' British operations got access. Rees, in an interview with Reuters, said the clock is ticking.
"Now is the time for us to think very strategically about what we need to do to protect our leading position moving forward. Time really is of the essence, we don't have two years here," Rees said.
The CEO of a major British bank told Reuters last week there is still no timeline for Mythos access. An Anthropic spokesperson said the rollout has begun outside the U.S. but remains coordinated with Washington.
"We have begun the rollout of Mythos 5 to organizations outside the United States. We continue to coordinate with the U.S. government to expand access to the broader set of domestic and international partners," the spokesperson said.
Rees and co-adviser Rohit Dhawan, head of AI at Lloyds Banking Group, published a package of recommendations on Tuesday as part of the government's financial services AI adoption plan. Their proposals urge regulators to scrutinize AI chatbots giving financial advice and to designate AI firms as "critical" providers, bringing them under direct oversight.
Finance minister Rachel Reeves said in a statement she has "set out a serious plan for AI sovereignty, backing British companies to compete and win."
The Mythos snub is not an isolated incident but a symptom of a broader strategic deficit. The UK's financial sector, which contributes over £170 billion annually, relies heavily on a few U.S. tech giants for cloud and AI. The Bank of England has repeatedly flagged concentration risk, warning that an outage or policy shift at a single provider could cascade through the system.
Rees argues the answer lies in diversifying partnerships beyond the U.S., pointing to AI firms in China and France. Yet such a pivot carries its own geopolitical and regulatory baggage, particularly around data sovereignty and alignment with Western security frameworks.
Dhawan's proposal to designate AI providers as "critical" mirrors a move last week that brought four U.S. cloud firms under the Bank of England's supervision. Extending that regime to AI models would force transparency and resilience standards but could also trigger pushback from firms wary of operational constraints.
The recommendations land as the EU advances its own AI Act, creating a patchwork of rules that global banks must navigate. Without swift action, the UK risks being sidelined in the next wave of AI-driven financial innovation, where cybersecurity and real-time risk modeling become competitive differentiators.
Rees's urgency underscores a harsh reality: in the race for AI supremacy, access is power, and Britain is currently standing outside the gate.
