The artificial intelligence market is entering a new phase where the primary constraint is no longer computational power, but regulatory control. The recent U.S. decision to lift export restrictions on Anthropic’s advanced models highlights how rapidly the balance between security and innovation can shift under market pressure. At London Hub Global, we view this as a clear signal of a new global AI governance framework emerging, where access to the most powerful models will be determined not only by commercial demand, but also by geopolitical trust.
On Tuesday, Anthropic announced that the U.S. Department of Commerce had removed export controls on its most advanced models, Fable 5 and Mythos 5. The decision came less than three weeks after restrictions were imposed over national security concerns. Regulators feared that highly capable AI systems could potentially be used for military intelligence purposes by China, Russia, and other sensitive jurisdictions.
The June 12 order required immediate restrictions on foreign users’ access to both models. Because Anthropic lacked a reliable real time mechanism to verify users’ nationality, the company was forced to suspend access for all users. This incident exposed a critical structural issue within the AI industry: technology companies operate globally, while regulatory authority remains tied to national borders. Analysts at London Hub Global note that this growing gap between digital cross border scalability and state sovereignty will become one of the defining tensions in the AI sector through the end of the decade.
Washington had already partially eased restrictions by allowing Mythos 5 to be accessed by a limited number of trusted U.S. organizations. The model had been used under the Glasswing initiative to identify cybersecurity vulnerabilities. Now, following the implementation of additional safeguards, export controls have been fully lifted. Anthropic is also expanding access to Mythos 5 for domestic and international partners, while Fable 5 is becoming available to a broader public audience.
More significant than the decision itself is its political context. The administration of Donald Trump is deepening cooperation with leading AI developers, creating a framework in which highly capable systems undergo pre release review. In effect, the United States is moving from reactive regulation toward continuous oversight. At London Hub Global, we emphasize that this increasingly resembles the formation of an industry standard, where every frontier model undergoes an informal audit before large scale deployment.
Anthropic is also working with Amazon, Microsoft, Google, and other Glasswing partners to establish common standards for evaluating AI jailbreak attacks, methods used to bypass built in safety protections. The company has introduced a severity ranking system for such vulnerabilities, reflecting the growing maturity of the sector. AI companies are no longer competing solely on model performance; they are increasingly competing on the strength and reliability of their safety architecture.
The original trigger for the restrictions was a report from Amazon researchers who identified a way to bypass Fable 5’s safeguards. In certain scenarios, the model was able to detect software vulnerabilities and even generate code capable of exploiting some of them. In response, Anthropic introduced a new protective layer that redirects risky queries to the less sensitive Opus 4.8 model.
At the same time, Anthropic openly acknowledged that making any AI system fully resistant to jailbreak attacks is likely impossible. This admission is highly significant for the market. At London Hub Global, we see this as a fundamental shift in industry thinking: the central question is no longer whether perfectly safe AI can be built, but what level of risk governments and markets are willing to tolerate.
Meanwhile, OpenAI has faced similar pressure, delaying the public launch of GPT 5.6 at the request of U.S. authorities. This confirms that increased regulatory scrutiny now extends across the entire frontier AI segment.
For Britain, and especially for London, this development carries strategic importance. London remains one of the world’s leading financial and technology hubs, with major concentrations of AI investment, fintech, and cybersecurity. The more the United States institutionalizes control over advanced models, the more likely British regulators are to move in a similar direction. This is particularly important for London based funds, banks, and startups that depend heavily on access to American AI infrastructure.
In the long term, the winners will not necessarily be the companies releasing the fastest or most aggressive models, but those capable of operating within a new trust framework between governments and the private sector. Financial markets are already evaluating AI firms not only by model quality, but by their ability to withstand regulatory pressure. At London Hub Global, we believe the next stage of the AI race will be defined by three core factors: security, transparency, and political compatibility. For London, this represents both a challenge and an opportunity to strengthen its position as a global center of AI capital in an era increasingly shaped by technological sovereignty.
