At Washington’s request, Anthropic out of the blue took its latest and strongest AI fashions offline over the weekend. The American firm mentioned it had little selection after the White Home demanded it block entry for all international nationals, together with its personal workers. Overseas, the incident provided a sobering reminder that the US not solely dominates frontier AI — its authorities additionally wields energy over who will get to make use of it.
The Trump administration’s motion was swift, sweeping, and imposed with little warning or clarification. The unprecedented shutdown of the Fable 5 and Mythos 5 fashions — which have been already topic to safeguards limiting their use in “high-risk areas” — that adopted gave new drive to long-running arguments cautioning in opposition to counting on the US for vital applied sciences. It was recent ammunition for the politicians, governments, and corporations already arguing that they should lead within the expertise themselves.
Within the UK, AI and on-line security minister Kanishka Narayan didn’t point out Anthropic, Donald Trump, or the US immediately, however used the shutdown to argue that Britain should develop its personal AI capability, framing the difficulty as a matter of nationwide safety. “We deal with each different risk to our sovereignty with lethal seriousness, however we haven’t discovered to deal with this one in the identical manner,” he mentioned, as pictures of British police and army flashed on the display screen. AI is “the central political query of our time,” Narayan mentioned, arguing that Britain should resolve how the expertise will form its financial system, safety, and sovereignty “earlier than another person decides the reply for us.”
In France, the response was extra express — and extra forceful in naming the US. Former Prime Minister Gabriel Attal, the presidential candidate for Emmanuel Macron’s Renaissance social gathering, referred to as the shutdown the beginning of “the AI warfare” and mentioned it reveals France’s vulnerability if it depends on others for vital applied sciences. He likened the pullback of Anthropic’s fashions to Iran’s blockade of the Strait of Hormuz, with entry to AI now a strategic chokepoint for which France should put together. Attal is much from alone: Le Monde reported related alarm from throughout France’s political spectrum.
The argument shouldn’t be precisely new. Europe has spent years fearful about its dependence on the US, technological or in any other case, and the European Union has put rising emphasis on decreasing the area’s reliance on exterior suppliers in areas like chips, cloud computing, and AI. However the Anthropic shutdown has made issues really feel extra rapid, including to an already deep unease over America’s reliability as an ally beneath Trump, from commerce disputes to threats of withdrawing from NATO. Attal mentioned the difficulty will likely be on the coronary heart of France’s subsequent set of presidential elections, whereas members of the European Parliament have pointed to the withdrawal of Mythos and Fable as proof Europe must make tech sovereignty a actuality, and to take action rapidly.
Canada has drawn an analogous lesson to Europe. Prime Minister Mark Carney mentioned the scenario highlights the danger of counting on only one accomplice for entry to essential assets like AI. “The scenario we’re in collectively proper now with Mythos and Fable is one thing that may occur with overreliance on sure fashions,” he mentioned. “No person has finished something mistaken within the scenario. However we can have finished one thing mistaken if we simply settle for this, don’t take the lesson, don’t construct out and diversify.”
Others are nicely down that highway already. Beijing has lengthy championed home AI corporations, and China is without doubt one of the few locations with fashions able to credibly rivaling the merchandise of American frontier AI labs. Nonetheless, in some areas, Chinese language fashions do lag behind their American counterparts, and Anthropic has accused Chinese language rivals of utilizing its fashions to coach its personal on an “industrial” scale. A part of the White Home’s resolution to tug Mythos reportedly stems from its perception a bunch linked to China had accessed the mannequin.
Most governments and companies can not come near matching the dimensions and assets of frontier labs within the US or China. However sovereign AI doesn’t all the time imply constructing the largest or essentially the most highly effective instruments. France’s Mistral and Canada’s Cohere present that stable efforts can come from exterior these international locations, even when the fashions can’t stand toe to toe. Different international locations, like Singapore and the UAE, have targeted on narrower however nonetheless strategic priorities comparable to infrastructure, or fashions that work higher with native languages. After all, there are additionally open-source fashions that might in the future have Mythos-like capabilities that will be laborious for any single social gathering to regulate.
Trump may even see limiting Mythos and Fable as a matter of nationwide safety. However the argument cuts each methods, and with Washington now asking if AI is just too essential for everybody to have entry, different governments are asking whether or not they can afford for Washington to resolve who does.
Anthropic could quickly carry Mythos and Fable again on-line. Restoring world belief in American AI is one other factor solely. Regardless of how lengthy the shutdown lasts, it shined a lightweight on how fragile entry to US frontier AI fashions is. Many governments and corporations didn’t like what they noticed — and are fired up to verify it doesn’t occur once more.