The most advanced AI systems are starting to become restricted.
Anthropic has announced that global access to Fable 5 will be restored tomorrow, but likely in a limited form, with coding and debugging tasks being routed to Opus 4.8.
OpenAI also announced last week that it was delaying the release of GPT-5.6 at the request of the U.S. government.
Access is no longer just about paying for a tool.
It is increasingly about being part of a small group of approved clients, partners, or organizations.
This changes the rules of the game.
We are entering an era of pre-approval checks, country-based and user-based restrictions, and growing uncertainty for companies building on top of these models.
Fortunately, open alternatives continue to evolve.
They may not yet match the most advanced closed systems, but they are becoming extremely powerful, closing the gap fast, and can already be deployed on private infrastructure.
The real question is: for how long will open, private, and sovereign AI alternatives continue to exist?
Access to certain AI models may end up looking more and more like access to financial services:
official ID verification, selfies, full activity traceability, and little to no confidentiality or privacy.