Washington now has a say in who gets the newest AI models first
Washington now has a say in who gets the newest AI models first. OpenAI is limiting its new GPT-5.6 models to a 'small group of trusted partners' at the US government's request; Anthropic disabled two of its latest models to comply with an export directive, one of which the government partially un-blocked days later.
What it actually means
For the first time, 'can I use the best model?' has an answer that isn't about price or a waitlist, it's about whether a government will let it ship, and whether it can be switched off after it has. The practical read for anyone building on frontier models: the newest thing is no longer guaranteed to be available the day it's announced, and access can change after the fact. That's a supply risk you didn't carry last year. Don't build a product that only works on the single newest model from one lab, keep a tested fallback to the previous generation or a second provider, because 'the model exists' and 'you're allowed to call it' are now two different things. This isn't doom; most of you are building on models that are months old and perfectly capable. But if your roadmap assumes day-one access to whatever's newest, that assumption just got fragile.
'The model exists' and 'you're allowed to use it' are now two different things.
Source: CNBC · Tuesday, 30 June 2026