We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5.
We'll begin restoring access tomorrow, and will share an update soon.
We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models.
You'd be forgiven if you felt like "sovereign AI" has been a buzzword.
Except now it's a necessity. Not tomorrow, not in a year, now.
Two weeks ago, the US government took down Fable 5, and this week they're restricting the release of GPT-5.6.
It's not just Anthropic models, and soon it won't just be American models.
Foreign governments can and will restrict the export of their most powerful frontier models - even to their allies.
Home-grown models developed in Canada will be the only way we can keep up on the world stage.
I'm with Kim here. This doesn't sound very good.
OpenAI was founded in 2015 with one clear mission: ensure AGI benefits all of humanity and avoid concentrating power in the hands of the few. That was literally their founding charter. Broad distribution, not restriction.
Now we've reached the point where frontier models need government sign-off before release. GPT-5.6 is being rolled out case by case during preview. If the US government becomes the gatekeeper for what gets released and when, international users don't just face delays - they face the real possibility of permanent restrictions or zero access.
And I'm not just worried about the US getting priority access. I'm worried about the rest of the world losing access entirely. If frontier models require approval from governments that don't govern you, what happens when those governments decide certain regions, certain countries, or certain use cases don't qualify?
I've already downgraded my OpenAI and Anthropic subscriptions. I'm looking at Kimi and GLM as backups, possibly daily drivers if this becomes the status quo. Because if frontier intelligence becomes approval-gated by region or use case, we're watching the exact opposite of what OpenAI was created to prevent.
I don't buy the safety story. Not from the labs, not from the government.
Fable, gone. Mythos, gone. Then all of GPT-5.6, down to Luna, the cheap everyday model for summarizing and drafting.
A summarizer is not a cyber weapon. But cheap intelligence is the one thing that lets everyone else catch up. So they locked it down too.
It was never about security. It's about who gets the frontier, and who stays behind.
@MonaSohal You’re welcome! Hoping it will come back soon but beginning to lose hope for us in 🇨🇦. There are some design features I want to try it with….
@ICEzzyG To Detroit: Connor Hellebuyck
To Winnipeg: Sebastian Cossa + 2027 1st, top-10 protected + Michael Rasmussen or Jonatan Berggren + John Gibson if Winnipeg wants a bridge goalie