Anthropic’s latest troubles should give automakers pause

The China Problem™️ is incredibly nuanced, and there isn’t a simple social media post-sized solution. Nor is it a problem I intend to solve today in this newsletter. (Though I’m certain my solution makes the most sense). Instead, if I want to talk about Anthropic.

Last week, Anthropic — the creators of Claude — made its most advanced model available to the public. This model comes in two variants, Fable and Mythos (depending on what type of customer you are), and is apparently really good at advanced tasks.

Are those tasks good enough for script kiddies to construct malware? Apparently. But that isn’t even why people should be concerned. What should be concerning is what the United States government did on Friday evening. At 5:21 PM Eastern Daylight Time, Anthropic received a directive from the government prohibiting any foreign national from using Fable 5 or Mythos 5. It didn’t matter where this foreign national was at the time, or even if that person worked for Anthropic. The government said “nope,” and the only way for Anthropic to comply was to shut down everyone’s access to the model.

There are a couple of concerning things about this, in relation to the automotive industry. At any given time, the United States government can claim that a piece of software poses a risk to national security and effectively shut down its use. Not only that, but basically shut down use of that software anywhere in the world.

Weird, because I thought China was the threat with things like this? But I digress.

If our government can effectively shut stuff down like this with no real rhyme or reason, what motivation is there for automakers to use U.S.-sourced software? When we’re talking about cloud computing and infrastructure for connected vehicles, the U.S. has proven itself to be an unreliable actor.

With Anthropic, it appears to be the only one forced to limit access, even though Anthropic claims other models could also be used, as the government says Claude could be used. So it appears this could be as much punitive as it is an actual concern. Perhaps Senator Slotkin should concern herself with this, too?

One of the difficult things for our elected officials to understand is that it appears like hypocrisy when they work to ban Chinese software and vehicles for national security reasons, but then our own government and our own private companies do things like this, or mishandle our personal and private data. There are legit national security concerns regarding China, but it’s hard to make that case when the world’s first trillionaire, enabled by our government and our policies, instigates rioting on the social network he purchased that exists to generate Nazi propaganda and CSAM. We, as a country, don’t need to be perfect, but we need to get our own house in order to protect the people, just as much as we need to be concerned about foreign actors.

There has already been a push in the European Union to stop using U.S.-based infrastructure. If I were an automaker, I wouldn’t use it. And if I were someone like Hyundai, who in the past couple of years has moved extensively to using AWS, I’d be demanding reassurances from Bezos and company that this type of thing wouldn’t happen to them.

That non-connected Slate truck might be on to something.