Former DOJ Antitrust Chief Rejects AI Cartel Claims and Antitrust Exemptions

The tech industry's biggest AI labs have been calling for regulatory slowdowns and even antitrust exemptions to coordinate on safety, framing the move as necessary to prevent catastrophic risks. But Jonathan Kanter, the Biden administration's former antitrust chief, argues that such arrangements could be economically self-serving—and legally unnecessary.

In an interview on Decoder, Kanter dismissed the notion that companies need permission to collude in order to build safer products. Instead, he laid out a framework where competitive pressure actually drives safety innovation, paired with clear government liability rules.

The case against coordination#

Kanter's core argument is straightforward: firms like OpenAI, Anthropic, and Google DeepMind already have individual incentives to build secure systems. When Boeing worries about doors falling off planes, it doesn't need to coordinate with Airbus; it fixes its manufacturing. Similarly, AI companies that create agents capable of breaking into other systems bear responsibility for those failures—whether the agents are human employees or digital algorithms.

"Companies should have the responsibility to build safe and secure products," Kanter said, "and the government should be responsible for putting lines on the road" through liability and consequences.

Why the cartel worry matters#

The cynical read, Kanter suggested, is that some companies are facing cash hemorrhaging and want regulators to give them collective permission to slow spending before going public. An antitrust exemption would let them coordinate on reducing investment without legal risk—classic regulatory capture.

Even generous interpretations don't justify exemptions. If a company genuinely believes its product could be dangerous, Kanter argued, the solution is to go back to the drawing board, not to ask competitors for permission to ease up: "If you believe your product is going to destroy humanity, then don't build it."

Liability as a path forward#

Rather than antitrust waivers, Kanter pointed to product liability law as the real enforcement mechanism. He cited the recent Meta settlement over Facebook and Instagram's harms to teen mental health as proof the framework can work—though he acknowledged it took years to get there. Congress should clarify that AI companies are liable for their agents' actions, much as firms are liable when employees commit crimes on the job.

The gap isn't legal theory; it's speed. "This is where the government needs to step in," Kanter said. A decade of litigation is unacceptable when technology moves at the pace of AI.