The recent Anthropic Fable 5 episode may turn out to be one of those moments we look back on as a shift in how businesses think about AI. For a short period, Anthropic was forced to disable access to its Fable 5 and Mythos 5 models after a US government export-control directive restricted access by foreign nationals. Anthropic said the practical effect was that it had to abruptly disable both models for all customers while it worked out how to comply. The restrictions have since been lifted, following additional safeguards, but the lesson remains. Access to AI models can suddenly be revoked, not because the model stops working or demand is too high but due to external factors beyond your control.
Until now, most businesses have treated AI models as a commodity. There might be limits on compute, queues, usage caps, price changes or occasional service disruption. But once a model was released, the assumption was broadly that you could get access to it when you needed it. That assumption no longer holds, which creates the question of what happens if you build important parts of your business around a particular model, provider or AI product, and then you lose access? If your internal workflows, customer experience, software development, reporting, automation or decision-making processes are built too tightly around one provider then losing access to that provider would be nothing short of catastrophic.
This is not just about Anthropic. It could equally be OpenAI or Microsoft, or one of the increasing number of products these companies offer like Claude Code, Design, or Co-Work, Codex, many of which are becoming embedded in the way a business works. The trigger could be anything from regulatory changes, safety concerns, export controls, commercial disputes, pricing changes, product decisions or technical constraints. We already knew AI costs and roadmaps were uncertain, now we also know that access to the model itself is too.
Your business needs to use AI to be successful in 2026, but is not enough to say “We use both OpenAI and Anthropic, so we are covered.” That does not solve the deeper problem. The right response is to become more produce and model-agnostic, designing your own AI systems where models can be swapped in and out, running alternatives in parallel where it matters. It means understanding which workloads require frontier proprietary models and which can be handled by cheaper, smaller or open-source models. It means avoiding unnecessary dependence on a single interface, a single vendor or a single product wrapper.
This is a form of AI sovereignty – not in the geopolitical sense, but in a practical business way. You need to know how your AI systems work, where the dependencies are, which models are being used, what fallback options exist, and what would happen if access suddenly changed. The businesses that get the most value will not simply be the ones that buy a licence to the latest tool but will build the right capability around the tools.
Some of that capability may be internal. Some of it may need to come from a partner. But the principle is the same: do not build your business on a foundation that can be pulled from underneath you. Because in the age of AI, vendor dependency is not just a technical detail, it is a board-level risk.
