Josh Hawley, a Republican, is a U.S. senator from Missouri.
Former artificial intelligence researcher Jacob Coxon recently told the world that the companies he once worked for are driving humanity to the brink of extinction. Days after he made his claims, tech CEOs seemed to agree, and in a series of lengthy manifestos and X posts, asked for … an antitrust exemption. Wait, what?
This is where we are. AI companies and their employees alike are loudly proclaiming the unprecedented dangers of the very products they are designing — while frontier labs demand unprecedented special treatment from government.
It’s time to change the game. AI may well be a threatening invention, but it is most certainly a powerful one. And the companies that are pioneering it have already amassed incredible power — influence over the economy, over workers, over politics and increasingly over the lives of everyday Americans. What these companies need is not more power, to say nothing of a government handout, but some long-overdue accountability. One way to give it to them is to let the people they harm sue them in court.
Whatever the truth of the doomsday scenarios, the latest AI products present real risks to real things Americans rely on — our financial system, hospitals, critical infrastructure. It’s no longer theoretical. During the recent Hugging Face cyberattack, a swarm of roughly 700 AI agents from OpenAI escaped a testing sandbox and hacked another company’s platform, then attempted to erase the evidence of their crime. And that wasn’t an isolated incident. We learned just days ago that other American AI agents hacked the government of Australia.
What happens when one of these hacks crashes an emergency room or shuts down people’s bank accounts? Who will pay?
The companies don’t want it to be them. And right now, the odds are stacked in their favor. Under current law, it’s not clear who would be responsible in these scenarios, if anyone. Tech titans are already suggesting their agents are acting “autonomously” and beyond their control. Meaning, it’s not the companies’ fault.
We shouldn’t accept that for an answer. I am introducing legislation in Congress to make the corporations liable for the actions of their agents when they design the agents in a reckless fashion. My legislation would also make users liable when they deploy the agents in a reckless way.
And it would toughen criminal penalties. The federal government already has laws on the books against hacking. It’s time to make clear those laws apply to AI companies. My legislation would give prosecutors the power to charge companies with crimes when the firms know their agents are capable of criminal conduct and fail to create reasonable safeguards.
We should apply the same rule to AI users. If a user knowingly deploys an AI agent and allows it to commit crimes, like crashing systems and stealing data, the user should be criminally liable.
This plan builds on a fundamental principle of American law: You are responsible for the damage you cause. If you build an AI model or deploy an agent that can cause havoc, then you shouldn’t be able to push the costs onto somebody else when bad things happen. Right now, the major AI developers — Meta, Anthropic, OpenAI and Google — have every incentive to move fast and break things, as Meta CEO Mark Zuckerberg infamously said, because they’re confident whatever they break will be paid for by the victims. We need to change that equation. (The Washington Post has a content partnership with OpenAI.)
If my bill passes, the incentives would shift. The labs would know that if they’re reckless, they’ll be the ones on the hook. Consequently, their behavior would change. They’d give more thought to protecting Americans’ rights, to protecting their data and to keeping their AI agents under control.
How can I be sure? Because corporate behavior changes every time you give power to the people to enforce their rights. Just ask asbestos manufacturers, who quit selling their cancer-causing products after they had to pay victims in court. If a private citizen or company can sue AI titans when they are hacked or crashed or otherwise harmed, something similar will happen. The companies will start internalizing risks rather than off–loading them on others. They will start behaving with the American people’s interests in mind, not just their profits.
There are plenty of other ways to take AI companies up on their request for federal rules. The Trump administration has rightly encouraged frontier labs to share their technology with the government before deployment to the public, so the models can be tested for national security concerns. We ought to make that testing regime mandatory. Congress can and should protect children from AI chatbots that push sexually explicit material or coach kids into self-harm. We can protect authors and creators against blatant IP theft and ensure data centers pay the costs of their gigantic electricity use so ratepayers don’t have to foot the bill.
But the key to it all must be a clear assignment of liability. That’s how we guarantee the costs of AI development fall on the people who design the technology, not on the working people of this nation.
In his warning earlier this month, Coxon claimed AI giants are “gambling with our lives.” As we assess the risks of AI, we should watch who is profiting off the panic. Tech titans should not get special treatment. If they want to gamble, they alone should pay the cost.
The post Josh Hawley: AI companies shouldn’t get a free pass to break things appeared first on Washington Post.
