One aspect of the Fable 5 shut down people are missing: Dario told the DoD to go pound sand and they are the only frontier lab to refuse to cooperate. Anyone worried that the same shut down will happen to OpenAI, who works with the DoD, is ignoring that political reality. This was a targeted hammering of the nail sticking out.
The Fable 5 shut down was likely planned well in advance based on the timing. It was just enough time for the model to earn hype and eyeballs, that way the hammer drop would also hit with maximum attention. It's to make an example of Anthropic: you either join the Military Industrial Complex, or you will not be allowed to release frontier models. And not just them, but every company, including those that will come later.
People are saying Anthropic's economic influence means they have power here. This is incorrect. The Fable 5 block means every company that relies on Anthropic now has to worry about the government shutting down their tools at anytime. This introduces uncertainty and companies hate uncertainty.
Investors also hate it, which directly impacts Anthropic's planned fall IPO. Why buy the stock of a company the government could dampen at anytime? We've just seen a very public mafia style kneecap on Dario to which he has very limited responses.
It's not the first time this kind of pressure has been applied to a company, either. And the pattern is the same every time: either they fall in line or they cease to exist.
Here's a prediction for next week or so:
Now that OpenAI has assessed Fable 5's capabilities, you will get GPT 5.6 that will significantly narrow the gap or may even close it completely. On release, the Dept. of Commerce will release short statement specifically saying that OpenAI complied with all requirements and that no action will be taken against them.
This will be the signal to companies and investors that OpenAI is safe and stable, while Anthropic is risky and uncertain. The market impact is obvious. It's entirely likely Altman worked out this strategy with government officials in response to Glasswing and that hype. (I have no specific evidence of this beyond Altman's prior contacts with administration officials and his cooperative attitude.)
It follows a pattern of behavior similar to Altman "buying all the RAM" to starve competitors. While I intensely dislike Altman's snake oil salesman tactics, he has generally been more politically astute than Dario. Outmaneuvering him like this is consistent with what we saw with the compute outcomes OpenAI had over Anthropic.
Dario can then either choose to play ball or see his company regulated into last place by DoD fiat until it is no longer a player.
The real problems come in when you ask what happens when those in power don't just want DoD access for frontier models. What if they want to push more censorship? Insert specific propaganda? Report anyone who asks the LLM certain questions? All of that is happening now to a lesser degree, but this opens the door to some very nasty outcomes down the road.