This kills the entire enterprise market for AI models better than Opus 4.8
1) No one is going to build any workflows/capabilities that could have the underlying intelligence rug pulled instantly by a bureaucracy or malevolent politician.
2) Even if a company was silly enough to take on the risk, is Anthropic going to ask all their enterprise customers to provide passports for all their employees and then setup individual Claude accounts for each and every employee of each and every enterprise customer in order to gatekeep access to Mythos? Because a plain ole api key no longer cuts it
This is the first impression I got, a glass ceiling on AI that will hit the market hard. Timing was right at the late Friday's let's-avoid-the-crash point, we'll see how this flies on Monday.
But it's also, as sibling comments mentioned, a bend-the-knee between Government and Anthropic. Once OpenAI catches up, and Anthropic lawyers-up as well, it will probably be reversed or morphed into a "models must have the US seal of AI-approval and, therefore and hereafter, we AI-approve the new US-verified Fable 5.1" - which will coincide with a at-large deployment at the DoD, Pentagon, and friends.
Otherwise the Chinese will catch-up and, heavens forbid!
And what if the US government does the same for Opus or other models? No model is safe from being banned that way
For American AI models better than Opus 4.8.
The much-decried EU AI Act provides a safe, predictable regulatory framework to base ones AI development on. It provides legal stability compared to unpredictable arbitrary decisions by the US executive.
If AI companies have any sort of sense in them, they'd be well-advised to consider relocating to Europe.
The EU AI Act killed Mistral. It's not a stable or reasonable base.
> It provides legal stability compared to unpredictable arbitrary decisions by the US executive.
I don't think assuming the European commission will act rationally or stably on this is really a good idea, and I say that as a European...
It's all a spectrum. Considering the alternative...
> If AI companies have any sort of sense in them, they'd be well-advised to consider relocating to Europe.
Too late now. They wouldn't be allowed to relocate in the name of national security.