US government directive to suspend access to Fable 5 and Mythos 5

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.


I thought this was a bit interesting.

And I'm a bit skeptical of Anthropic's sincerity in the way they're approaching this.

I wonder how much of the "Gee shucks, that means we have to shut it off for everyone" is actually out of necessity to ensure compliance, vs. it being a case where they shut it off for everyone, and then sit back and wait while the administration's phones ring off the hook with angry monied interests pressuring them to reverse course and modify the directive.

In any case, it's an unwise move to intentionally hamstring a domestic company when there's a global AI race.

durangodawood said:

As for gatekeeping for citizen status, I guess it depends on how strict enforcement would be and what the penalties are. Sounds hard for a private company to be that certain of the identity of all possible account holders.

Click to expand...

That same expectation was foisted upon US social media companies with regards to laws that Canada and Australia were trying to pass, and I kept explaining why that was neither feasible nor practical. (I think you may have been in those threads, for some reason, I tend to confuse you and Rambot in terms of past thread interactions)


But my honest answer to that is, the more implausible it is to implement from the corporate side, the more impossible it'll be to detect and prove on the regulatory side. If Anthropic has no way of being 100% sure if the person is an American citizen that's using it (even with ID verification requirements), the battle to prove that someone wasn't a citizen (if they past those safeguard checks) would be an even bigger uphill battle for the enforcement side.

In order for the US government to be able to catch an Anthropic mistake or loophole in their authentication process and prove an instance of it occurring, that would mean they'd need to be better at this than Anthropic... which they're not. If they were, they'd be writing their own superior solution in-house rather than trying to contract Anthropic.