Innovation demands integrity.
On Friday, the federal government forced Anthropic to take two of its newest and most capable models, Mythos and Fable, offline.
The order came from the Department of Commerce, which ordered Anthropic to block access to the models to foreign nationals as an emergency export control, claiming that their capabilities presented a security risk. As Anthropic noted, the difficulty of tailoring blocking access only to foreign nationals (including Anthropic’s own employees) meant that it had to block access for everyone.
This treatment is not consistent with the way other frontier models are treated, which cybersecurity experts have noted can be just as capable. Instead, it appears to be an extension of the administration’s ongoing targeting of Anthropic, particularly in the context of the company’s expected IPO. Anthropic is leading today, but what happens if investors believe that the administration is going to block or delay the company’s leading models?
This story has played out dozens of times already, beyond Anthropic and AI and across the economy. Federal law is full of emergency authorities, powers the government can use when ordinary procedure would be too slow or restrictive to meet urgent problems. By their nature, these authorities have to be broader and more open-ended than laws typically are; if they were too specific, they couldn’t serve their purpose. We rely on the integrity of leaders not to abuse them, but this administration sees them as backdoors to unchecked authority.
That’s a problem for American AI leadership. AI develops rapidly, but rapid growth is grounded in long-term investment. Companies need to be able to make long-term bets on building infrastructure, on designing chips, on training models. Researchers and engineers need to make long-term bets on the problems they’ll tackle and the institutions and companies where they’ll work. Investors, of course, have to decide where to put their capital. Long-term investment depends, in turn, on predictability and the rule of law — clear rules that make it possible to assess the risk and reward of all those bets.
When we think about American economic strength, we tend to think about the companies that put economic value into the world. But just as important are the legislators and regulators who set clear rules, the judges who enforce them impartially, and the officials who wield their authority with integrity and restraint.
By using regulatory powers for retribution and coercion, the Trump Administration is eating away the core of American economic strength. Even one action like this could be devastating for a company like Anthropic, but it’s the pattern that weakens the economy. First, it creates uncertainty that weakens investment. Then, companies start putting pleasing the administration at the top of their priority list. The winners will be the best sycophants, not the best innovators and operators. That’s no way to compete.
The irony is that this administration was put in place in part by tech-industry money, on the view that a Democratic administration would over-regulate and stifle innovation. There was something to that, but what those donors chose instead was chaos and cronyism. If they’d been better students of history with an ounce of humility, they would have understood that private-sector freedom to innovate depends on public-sector integrity.
Instead, the weakness of the President’s character is being translated into weakness in the American economy.