Cameron Otsuka

American AI Suzerainty

The Anthropic export control saga was not only a preview of AI safety policy, but also a peek at what a global reordering of political economy around AI intelligence could become. Access to frontier model intelligence, and possibly Powerful AI as Dario calls it, exhibits many of the same properties that other arms-/export-controlled technologies do: a thin line separating civilian and military use. Think rockets, cryptography, or satellite imaging.

The important escalation in the Anthropic case is that export controls moved from physical components, digital artifacts, and intellectual property like Anthropic’s Mythos model weights, which represent massive levels of fixed cost. Instead, controls were placed on access to the intelligence itself because militarily, Mythos is akin to a weapons R&D lab full of scientists. The diffusion of stronger levels of intelligence is irreversible as it’s an R&D proliferation event. As The Verge put it, “Anthropic got hit by export rules nobody understands.”

One possible outcome in this future world is the gradual march towards feudal states reliant upon good standing with the US. Allying with the US would enable access to superhuman research capabilities, cybersecurity, etc., whereas falling out of the privileged zone would place foreign countries at a disadvantage.

The UK and other G7 nations were abruptly made aware of this weakness and Japan’s Digital Minister has already hinted at this, starting conversations with other countries to develop their own sovereign AI capabilities:

I hope many Japanese people understand that we need to press ahead with AI development, or we’ll end up becoming an “AI colony.” — Japan Digital Minister, Hisashi Matsumoto

Access to Powerful AI becomes a magnetizing force into the US’ orbit, then. Becoming an American citizen, or headquartering a company within the US’ sovereign borders offers outsized benefits to being anywhere else in the world¹.

This is where the Yarvinite Patchwork thesis falls apart². While Patchwork assumes citizens’ ability to exit would force governance discipline, if Powerful AI is not widely diffused, then the market for city-states consists of a duopoly with a long tail of penny stocks.

While AI export controls won’t and can’t preserve the prior global order, they increase the US’ leverage in geopolitical negotiations, intensify the demand for domestic US labor/capital, and force foreign states into a choice between sub-frontier model intelligence and deploying massive amounts of their own capital to (hopefully) build sovereign frontier AI of their own.

Mine Print Hash

The Kevin Warsh era of the Fed has begun, bringing with him a slew of institutional changes already that Matt Dines talks through. I cover the Anthropic export controls and what they could mean as countries continue their push for AI sovereignty.

Open Threads

Can smarter model routing substitute for raw model intelligence? My guess would be no, but it could marginally improve performance?

• Japan’s Sakana AI claims it’s Fugu model can do this. Link
• OpenRouter claims it’s Fusion tool can do this. Link

Europe trying to sort itself out.

• UK chews through it’s 6th PM in 10 years as Starmer resigns. Link
• Hoping to update their monetary rails to be competitive with the US. Link
• Trying to get satellite launch capabilities going. Link