Yesterday I was on a panel at ENTREfest. The session was called Beyond the Hype: Panel Discusses AI for Entrepreneurs and Business Leaders. One of the questions was where companies should focus right now.
A couple hours after we finished, a good example of what to watch for showed up.
On June 12, the US government issued an export control directive citing national security. It blocked access to two of Anthropic’s models, Fable 5 and Mythos 5, for any foreign national, anywhere, including Anthropic’s own foreign-national employees. To comply, Anthropic shut both models off for all customers the same day.
Anthropic reviewed the government’s claim, said the flaws were minor and already present in other public models, and disagreed with the order. The models went dark regardless.
If you were running a product or an internal workflow on either model, calls to it started failing that afternoon. No warning, no transition window. You didn’t do anything wrong, and neither did your vendor.
That’s the part most businesses don’t plan for. This wasn’t an outage or a price change or a deprecation notice with months of runway. It was a regulatory order, and meeting it meant pulling the models.
Most businesses now have AI somewhere in their operations. A support chatbot, a document summarizer, a coding assistant. A lot of those run on one specific model from one specific company.
That’s a single point of failure. The list of ways it can fail now includes a regulator restricting the model, which isn’t something you can control or predict.
On the panel, someone asked what practical step business owners should take in the next month. Here’s mine. Look at where you depend on AI, and make sure you can switch the model out without rebuilding anything.
Keep the part of your system that talks to the AI separate from the rest. Pick a backup model you’d move to. Then a day like June 12 is an inconvenience instead of an outage you can’t fix.
