91探花

European businesses have spent the last few years integrating AI into different parts of their systems. Many selected the most advanced models available and built important workflows around them.

That decision looked sensible when AI providers competed on performance and usability. Few organisations expected access to those models could disappear because of a government order issued in another country.

Last week, Anthropic announced that a US government directive required the company to suspend access to its Fable 5 and Mythos 5 models for all foreign nationals. Anthropic said the order arrived at 5:21pm ET and required immediate action.

The announcement affects more than one AI company. It places ownership, jurisdiction and control at the top of the agenda for enterprises that depend on advanced AI systems every day.

听

What Happened At Anthropic?

听

According to Anthropic, the US government ordered the suspension of Fable 5 and Mythos 5 under national security authorities.

The company said the government believed it had become aware of a method of bypassing, or 鈥渏ailbreaking鈥�, Fable 5. Anthropic reviewed the demonstration and said it involved a small number of previously known vulnerabilities.

Anthropic defended its security measures and said extensive work had taken place before the models were released. The company said, 鈥淲e stand by this defense in depth strategy. It reduces the risks posed by Fable, making them comparable to the risks of existing models already deployed across the industry.鈥�

The company also disagreed with the decision. Anthropic said, 鈥淲e are complying with the government鈥檚 legal directive and are removing access to Fable 5 and Mythos 5 for all users. However, we disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people.鈥�

Anthropic apologised to customers and said it was working to restore access.

听

Why Are European Enterprises Worried As Well?

听

Many European organisations built AI systems around US hosted frontier models because those models delivered the performance businesses wanted.

Tim Sears, Chief AI Officer at HTEC, said many enterprises never seriously considered the possibility that access could disappear overnight.

Sears said, 鈥淭he Anthropic situation is a wake-up call for any European enterprise that has built critical workflows on a single US-hosted frontier model, and the uncomfortable truth is that most firms have.鈥�

He believes those decisions made sense when they were made. Sears said, 鈥淭he technology decision felt safe at the time because the models were better, the tooling was more mature, and nobody was seriously asking what would happen if access were pulled overnight.鈥�

European organisations have spent years examining where their data is stored. According to Sears, much less consideration went into where AI inference takes place and which legal jurisdiction controls the model itself.

Sears said, 鈥淲hat makes this particularly significant is that it reframes a conversation the industry has been having only partially. Data sovereignty regulation has pushed enterprises to think carefully about where their data sits, but where inference runs and under whose jurisdiction the model itself operates has not been treated with the same seriousness.鈥�

When it comes to enterprises operating in regulated sectors, that issue now has immediate business relevance.

听

听

What Could This Mean For Financial Institutions?

听

Banks, payment providers and financial institutions depend on software systems that must operate reliably and securely.

Azimkhon Askarov, Co CEO and Partner at CONCRYT, believes the Anthropic suspension deserves attention throughout financial services.

Askarov said, 鈥淭he recent news that the latest Anthropic Claude Mythos AI model has been suspended due to security concerns should raise alarm bells for major financial systems around the world.鈥�

He said the issue reaches into banking infrastructure, payment networks and the systems that support economic activity.

Askarov said, 鈥淲e鈥檙e ultimately talking about the resilience of the banking system, the exposure of core operating systems and payment networks to bad actors, and the erosion of trust in the financial systems that underpin the global economy.鈥�

Askarov also believes responsibility cannot rest entirely with regulators.

He said, 鈥淩egulation will inevitably play a role, and rightly so, but it won鈥檛 keep pace with the speed at which AI is evolving. That means businesses can鈥檛 afford to wait for policymakers to provide all the answers.鈥�

He added, 鈥淭he responsibility sits with the organisations building, deploying and benefiting from these technologies today.鈥�

听

Where Do Businesses Go From Here?

听

Many enterprises are now reviewing how dependent they have become on a single AI provider.

The shutdown has also brought renewed interest in alternative models and local infrastructure options.

Sears believes enterprises have more options available than they did 18 months ago. He said, 鈥淭he good news is that the options available to European enterprises today are genuinely better than they were even eighteen months ago.

鈥淥pen source models have reached a level of capability that makes them viable for a much wider range of production workloads than most people鈥檚 mental model of them reflects, and on-premise inference is a real option in a way it simply was not before.鈥�