On June 12, something unusual happened. Access to two of the most advanced AI models, Anthropic’s Fable 5 and Mythos 5, was turned off like a water faucet.
The company shut down access after a U.S. government export control directive, citing security concerns, required it to block use by foreign nationals inside and outside the U.S. In a statement, Anthropic said, “The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 to all our customers to ensure compliance.”
That includes enterprise users who aren’t foreign nationals.
“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,” the statement said. “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”
That’s the story for the AI model providers. What does this mean about AI control for enterprises relying on these models?
For enterprises building critical workflows on top of AI, the message may be that access to intelligence can now be throttled or even revoked based on jurisdiction, not just technology.
Box CEO Aaron Levie wrote on LinkedIn, “If at any moment a model can be become unavailable to your country’s users or businesses, this poses very real risk on relying on technology from a particular country.”
That’s why sovereign AI—a country’s ability to develop, host, and control its own AI using its own data and infrastructure—is quickly becoming an enterprise concern.
If a core function such as customer support or IT operations depends on a model that can be restricted externally, what guarantee does any enterprise have that the AI models they rely on will always be available?
The bottom line is that most enterprises are adopting AI capabilities they cannot fully control. So how can enterprises mitigate risk and insulate themselves from disruption?
The answer probably won’t involve AI model selection, but rather changing where control over AI lives. The ServiceNow AI Platform, for example, governs any model, agent, or identity across the business.
Instead of tying operations to one provider, the platform discovers, monitors, and manages AI systems regardless of whether they come from internal development or third-party vendors. This gives organizations a single layer of visibility and policy enforcement.
When access to frontier AI models can be restricted in a blink, enterprises need a layer that sits above the model, where workflows can be rerouted in the event of disruption, risks can be governed, and rogue AI behavior can be shut down immediately.
ServiceNow’s approach centers that control in the workflow layer itself (including an all-important kill switch), connecting AI strategy, governance, and execution so that outcomes remain stable even when underlying models change.
The ServiceNow AI Platform connects systems, orchestrates work, and governs processes across the enterprise—work that extends to AI-driven workflows.
When AI is woven into workflows, the platform is the place where organizations can separate dependencies on models so that workflows can be rerouted or updated without disrupting work. They can also apply consistent governance across regions, model providers, and use cases.
AI is becoming a governed resource subject to regulation and geopolitical pressure, which should change how enterprises think about adoption and governance. How can you keep your business running when AI dependencies change based on things that are out of your control? The answer depends at least partly on whether you have a platform of discoverability, governance, and control.
The Anthropic episode demonstrates that organizations must control how AI is used, where it runs, and what happens if and when it fails.
Find out how ServiceNow can help you control and govern AI in one place.
