Slippery slop: Dangers of AI titans policing themselves

LINA M KHAN

Anthropic’s chief executive, Dario Amodei, has transfixed Washington with apocalyptic warnings that his products could end humanity. To prevent such an outcome, he has proposed letting the artificial intelligence companies police themselves.

The government’s role? To issue an antitrust exemption allowing the companies to coordinate on slowing down model development and try to persuade China to adopt a similar approach. And on Tuesday, President Trump sat flanked by top AI executives after a private lunch at the White House and promised, “tremendous self-regulation.”

AI may differ from products developed during the internet boom, but some parallels are emerging. Starting around 2018, Meta began self-governance initiatives to introduce some oversight of its behaviour. In practice, these measures seemed more intended to forestall government scrutiny. Almost a decade on, jury trials have revealed that Meta was concealing evidence that its products seriously harmed kids and teens.

Implicit in the idea of AI self-regulation is that America is incapable of managing risk through our democratic government. But historically we have done just that, successfully building and regulating powerful new technologies, from railroads to nuclear energy. We can bring AI companies in line with the public interest and deter dangerous behaviour.

First, policymakers should look to laws already on the books, not to the online manifestoes and blog posts written by tech chief executives, for guidance. Much of the commentary from the executives suggests they currently operate in a legal vacuum. In fact, their business practices — including how they develop and deploy their technologies — are already subject to longstanding laws.

Product liability laws apply when goods are defective or dangerous, while tort law establishes basic duties for companies to not act negligently or recklessly. Consumer protection laws prohibiting unfair or deceptive practices can also kick in when businesses injure others by failing to take adequate precautions.

A host of lawsuits in this vein have already been filed against AI companies, with several judges already affirming the viability of these cases at the initial stage. State attorneys general can also act. Individual executives at AI companies can also be held personally liable for unlawful activity, and we already have criminal laws that address more egregious forms of corporate lawbreaking. Given the automated hack of Hugging Face and a series of similar breaches, enforcers at both the state and federal levels should immediately hold AI executives accountable for the dangerous and illegal actions of their systems.

Congress will need to legislate. America has a long history of regulating companies and products that are both transformative and risky, including nationally chartered banks, potentially dangerous pharmaceuticals and lethal nuclear assets. Drawing on this tool kit, lawmakers should consider mandatory testing of advanced AI systems; breaking up the AI giants to prevent conflicts of interest that can arise when a dominant company controls a critical technology, its inputs and its distribution; and truly independent supervision to monitor risk.

Lastly, tackling the challenge of AI will require confronting our broader crisis of governance. The flood of money in politics greenlit by the Roberts Supreme Court has also distorted our legislative body, where even popular, bipartisan initiatives now fail to pass, or often fail to even get a vote.

Over this same period, our court system has become increasingly libertarian on economic issues, with judges regularly vetoing policies that would promote economic fairness and protect Americans from corporate abuse while going easy on corporate lawbreaking. Trump’s executive branch has devolved into naked corruption and stopped enforcing laws against wealthy interests.

Outsourcing regulation to the AI companies may prove tempting. But it would be a clear mistake. Our democracy cannot surrender the trajectory of these powerful technologies to a handful of private actors. We must ensure they ultimately serve public ends.

Emerging technologies have always created both opportunity and peril as society and business adjust to rapid change. The public is now awake to the reality of AI’s potentially transformational nature. But the most dangerous reaction would be for Washington to panic and hand the steering wheel to the very same companies and executives whose breakneck pace of development and financial conflicts led us here in the first place.

Lina M. Khan is an associate professor of law at Columbia Law School and chair of the Board of Directors of the N.Y.C. Economic Development Corporation, which promotes the city’s economic growth. She was chair of the Federal Trade Commission from 2021 to 2025