Neither Law Nor Order

What the war against Iran and the sanctions against Anthropic have in common

I would really like to write an intelligent post about the likely consequences of the two most important events of the last few days, the joint Israeli-American attack on Iran that began Saturday morning and the Pentagon’s blacklisting of Anthropic for rejecting their demands for total control of their model. But I have to be honest: the cone of possible futures from both actions extends far too wide for effective prognostication.

The express purpose of the attack on Iran was to decapitate Iran’s government, cripple its means of self-defense or retaliation, and thereby usher in a change of regime. The administration has said very little, but it has said that much. But “change of regime” encompasses a wide range of possibilities. Will the regime implode, and a relatively liberal and pro-American government emerge in its place, as those who cheered Ayatollah Khamenei’s assassination both in Iran and elsewhere dearly wish? Or will exiled monarchists swoop in and attempt to take charge? Or will there be a civil war between ideological factions, some of them aligned with different potential international patrons? Could we see outright disintegration of the state, with ethnic separatism in the Azeri and Kurdish northwest and the Baloch southeast, and military intervention by Turkey and Pakistan in response? Or, alternatively perhaps the Revolutionary Guard Corps will seize the reins of power directly, inaugurating a more radical and fanatical Islamist regime? Or maybe the current oppressive regime will survive largely intact, led by new leaders more committed to collective survival than to any larger ambition and determined to keep a lower profile going forward, and will enter into an agreement something like the one that Oman’s foreign minister claims was largely agreed to before the bombing began?

I suspect that President Trump’s actual preference is for the last of these possibilities, resembling as it does the post-abduction outcome in Venezuela. But all of these options strike me as plausible—and America will have exceedingly limited ability to influence the course of events that we have set in motion, since we won’t have any presence on the ground. We’ll mostly just have to wait and see, and perhaps occasionally bomb someone we don’t like if they seem to be getting ahead.

And that’s only listing some of the possibilities for what will happen within Iran itself. Iran’s retaliatory attacks have pushed the Gulf States even closer to America for the time being, but that could quickly change depending on the trajectory of the war, including the scope of retaliatory violence, but also based on what happens politically inside Iran and how whoever next takes power is perceived. But beyond the Gulf, what will the impact be on the internal politics and external posture of nuclear-armed Pakistan? Or of NATO-member Turkey, the country so far to have benefitted the most from the collapse of Iran’s soi-disant Axis of Resistance? Or consider the effect on Israel and its relationship with the United States. Right now, the war is a sign of just how closely America is willing to work with Israel, fighting side by side in an aggressive war of choice. If the war goes well, that relationship could be deepened, Israel’s relationship with the Gulf states strengthened, and even its relationship with Europe somewhat repaired. But if things go pear shaped—and particularly if America suffers significantly from Iranian retaliation—public opinion, already less sympathetic to Israel than to the Palestinians for the first time ever, could turn very sharply against the Jewish state. Beyond that, I note that Israel’s once and quite possibly future prime minister Naftali Bennett is already warning that Turkey poses an increasingly dangerous threat to his country. Success in the war against Iran could breed attempts to run a similar play against a very different category of opponent, by Israel—or by the United States.

So let’s look through the widest lens. Iran has been increasingly tied to Russia in recent years, supplying them with drones for their war in Ukraine and sharing an ally in Assad’s Syria. The Syrian regime is gone now, and just as it got no support from Iran as its own military was collapsing, Iran apparently reached out to Russia this weekend, seeking support, and came back with nothing. Some are already speculating whether the Putin regime would prove brittle if the Russian public decided to take inspiration from events in the Middle East, but by the same token it could prove similarly ruthless in its determination to stay in power, and reasonably paranoid about what America might be planning for it. But what is America planning for it? And what more fantastical plans might we concoct if we decide that decapitation went swimmingly in Venezuela and Iran? And then there’s China. Xi Jinping probably has fewer worries about internal subversion than Vladimir Putin does (in part because he has been paranoid enough to give Khamenei a run for his money), but the loss of Iran as a functional power does remove one more contingency from American war-planners concerned about Taiwan. But it does so by drastically depleting America’s weapons stockpiles in the short term. Does that combination make an attack across the Strait less likely? Or does it accelerate the war’s timetable?

The point being: the United States is once again behaving unmistakably like a revisionist power, seeking to overturn the existing order, shaky as it was, by force. We’ve buried it a bunch of times already, but surely now we can say that the so-called rules-based international order—always substantially fictional—simply no longer exists. If we were operating in the realm of law, even the quasi-law that is operative between states, this war would unquestionably be illegal. But we plainly aren’t operating in the realm of law anymore. The best evidence of that conclusion is that countries like Canada and Germany—who we have grown used to thinking of as key believers in that rules-based order—have been the most vocal supporters of the American-Israeli attack.

Now, one can legitimately point out that the government of Iran flagrantly violated every rule in every book for its entire history. They routinely undermined legitimate governments in Lebanon and Iraq by funding terrorist groups that those governments could neither cow nor control. They never distinguished between legitimate military targets and civilians, and still don’t—they’re firing missiles and drones at whatever they can hit: hotels, oil refineries, synagogues, caring only that they are located in the territory of their adversaries. They have made assassination a regular tool of foreign policy, including attempting to assassinate President Trump before the 2024 election. And of course the regime rules without a shred of consent from their own people, having never allowed a free election and having repeatedly arrested, tortured, killed and most recently massacred tens of thousands of anti-government protestors. Why should the United States play by Marquess of Queensbury rules against a regime like that?

But the premise of living according to the rule of law is not that no rogue actors will ever gain advantage by breaking the rules but that those who do abide by the rules benefit from being able to trust each other, and that this benefit vastly outweighs the cost of having to deal even with rogue actors in a mostly rule-oriented manner. That’s a premise that this administration simply rejects. They not only don’t care about the U.N. charter or the other rotten props of liberal internationalism—they don’t even pretend to be negotiating in good faith when they enter into negotiations. They are not interested in building trust; they are interested in building fear.

That applies domestically as well as internationally—and I’m not just referring to the way the war on Iran was launched without any authorization from Congress or any attempt to make a public case for their actions. This case is more egregious in that regard than any I can recall, but the imperial presidency has been built from manifold precedents of a similar if lesser sort. The other major event of the past few days—the blacklisting of Anthropic—is more dramatically precedent-setting, and exceedingly alarming as such on this question of whether we’re still playing by the rules.

The problem the government had with Anthropic is that it was building guardrails to keep its A.I. properly aligned that did not themselves align perfectly with the desires of the military to use those tools as it sees fit for whatever contingency might arise. The official sticking points were prohibitions on mass surveillance of American citizens and on fully-autonomous weapons systems, prohibitions which the Pentagon wanted removed even though they claimed they had no intention (or authority) to use the system for either purpose. But I think their larger concern is that A.I. is a black box that the Pentagon simply couldn’t be sure wouldn’t fail to perform as they want it to under some unknown future scenario. I don’t think this is a solvable problem, because A.I. is a black box, and unknown future scenarios are, you know, unknown, so you really can’t possibly assure yourself that an A.I. will perform the way you imagine you’d want it to in every possible scenario. That doesn’t mean that I don’t think the problem isn’t a serious one, of course; lots of serious problems can’t actually be solved. Regardless, I do understand why the military wouldn’t want a private company structurally inside its decision making loop under any circumstances.

Given their concerns, the logical thing for the administration to do in response would have been to cancel its contract with Anthropic and turn to xAI or OpenAI or Gemini (all of which are now approved for use inside the military) as a replacement. Instead, the administration took the extraordinary step of announcing that Anthropic was a supply-chain risk, which meant that not only wouldn’t the military be using their products but that no company that does any business with the military could use them. In effect, the administration was saying that if Anthropic didn’t comply with the administration’s demands then not only they wouldn’t use it, but nobody would; its continued independent operation in any area that touched the government was a potential national security threat, much as the government views Huawei’s technology as an inherent security risk because of that company’s relationship with Beijing. Huawei’s market share in the United States, it should be noted, is roughly zero. This was a blunt assault on Anthropic’s independence and even its viability, and essentially unprecedented as such.

Now, once again, you can say that A.I. really is different from other products precisely because it acts as an agent, and that therefore every arrangement is essentially without precedent. But even if you take that view, we should still be open-eyed about the fact that the United States government is saying that if you build a frontier A.I. model and are subject to American jurisdiction you will effectively become an arm of the state. And given that these frontier models are expected to become pervasive in private industry, that means that in the very near future the state will be, effectively, inside every private company’s decision loop. That’s how I’d expect things to go in the People’s Republic of China, not in the United States of America—but that, again, is the point. We’re joining them because otherwise we aren’t convinced we can beat them.

Anthropic is going to take the administration to court, and perhaps it will ultimately win on the specific question of whether the supply-chain risk designation was legally proper (assuming the case isn’t settled long before it comes to judgment). But the administration has other tools to bring the company to heel, like the Defense Production Act, that would be harder for Anthropic to fight, and the mere fact that they would consider using them will likely shape the markets and corporate behavior, including at Anthropic. I don’t see how the next administration will be able to unring that bell, even if they want to—and I strongly doubt they’ll want to. So once again, the implication is that we are no longer really within the realm of law, and once again, the broadest cost is in terms of trust. The point of having the source of law follow the law is to foster trust generally, not only of the state but between people who don’t know each other. When you move outside the realm of law and trust, and into the realm of power and fear, that’s no longer really possible.

So regardless of whether the war in Iran “works out” in the sense of leading to at least a somewhat less terrible regime there and minimal American and allied casualties, and even if the sanctions against Anthropic “work out” in the sense that the company isn’t driven out of business and the Pentagon doesn’t wind up powering up Skynet—even if we avoid the worst, we’re entering a new world. And I’ve got a bad feeling about the new world we are entering.