Overview of Does AI need an antitrust exemption so it doesn't kill everyone????
In this Decoder episode, Neil Patel interviews former DOJ antitrust chief Jonathan Cantor about the growing push to regulate AI, the debate over whether frontier AI companies need an antitrust exemption to coordinate on safety, and how competition policy should be applied to AI labs. Cantor argues that AI does not need a special exemption to be made safer: companies can and should build safer products within existing law, while government should set clear rules, liability standards, and consequences for dangerous systems. He also frames the AI moment as a broader shift in U.S. politics, antitrust, and industrial policy, with China frequently used as both a real concern and a convenient justification for consolidation.
Main Themes
AI safety does not require a cartel
- Cantor rejects the idea that AI companies need an antitrust exemption in order to collaborate on safety.
- He says there are legitimate forms of cooperation, like threat-sharing clearinghouses and security repositories, that existing antitrust law already allows.
- But coordination to slow competition or reduce spending would look much more like cartel behavior.
Companies should be liable for harmful AI behavior
- Cantor argues that if an AI agent breaks into another system, steals data, or causes harm, the company behind it should be responsible.
- He treats AI systems like employees or tools under a company’s control: if they do bad things on behalf of the firm, liability should attach.
- He sees this as an extension of products liability and broader corporate responsibility.
Regulation is behind, but not impossible
- He says Congress has failed to create modern safety rules because the political system is broken and captured.
- Still, he insists the absence of federal action is not an excuse for companies to ship dangerous systems.
- In the meantime, courts, state laws, and products liability may have to fill the gap.
Antitrust, Competition, and “Rules of the Road”
AI needs basic guardrails, not a special carveout
Cantor uses a road-safety metaphor throughout:
- AI is like inventing cars and trucks without traffic lights, lane markings, or speed limits.
- The solution is not to let competitors coordinate freely; it is to establish rules for safe deployment.
Existing law can already handle much of this
- He argues Congress should clarify liability and set standards for:
- mental health impacts
- competition
- authenticity of information
- copyright and intellectual property
- He thinks the right goal is to force companies to build safer products from the start, not excuse harmful systems after the fact.
Strong antitrust still matters
- Cantor defends the Biden-era antitrust push, saying it changed the legal landscape and made competition policy more serious again.
- He says the goal is not to weaken big companies for its own sake, but to preserve room for challengers and prevent markets from hardening into monopolies.
China and the “National Champions” Argument
China is not a good excuse for monopoly
- Cantor pushes back hard on the idea that the U.S. needs domestic AI monopolies to compete with China.
- He argues the real way to beat China is to:
- keep U.S. markets competitive
- build safer systems
- protect users
- preserve freedom and innovation
- He says “national champions” may make sense in China, but they are contrary to the American model.
Foreign policy and competition law should stay separate
- He distinguishes national security concerns from antitrust.
- If Chinese models pose actual cyber or intelligence risks, those are geopolitical and security problems.
- But he says those concerns do not justify tolerating monopolization at home.
Political Realignment Around Big Tech and AI
Old alliances are breaking down
- Cantor notes the strange new alignment where:
- some libertarians are echoing Lena Khan’s anti-monopoly views
- some on the left and right now agree big tech is too powerful
- prominent figures from opposing camps are suddenly sharing anti-monopoly rhetoric
- He sees this as evidence that the old Reagan-era and neoliberal consensus is collapsing.
Public trust in tech is low
- He says tech firms have hurt themselves by openly predicting massive disruption:
- they say AI may kill jobs
- maybe even threaten humanity
- yet still push giant valuations and infrastructure buildouts
- That has made the public skeptical and more open to regulation.
What Cantor Would Do
His policy priorities
- Clarify liability
- Make it explicit that dangerous AI behavior carries consequences.
- Set values-based standards
- Protect mental health, competition, authenticity, and intellectual property.
- Allow safe collaboration
- Permit security sharing and safety coordination that doesn’t suppress competition.
- Preserve competition
- Keep markets open so new entrants can challenge incumbents.
- Avoid AI exceptionalism
- Don’t give AI companies special treatment just because the technology is powerful or geopolitical.
Notable Takeaways
- “We’ve invented cars and trucks, but we have no lines on the road.”
- “If you believe your product is going to destroy humanity, then don’t build it.”
- “The way we win against China is to make sure that we are delivering great innovations and products that benefit society.”
- “The challengers of yesterday are often the monopolists of today.”
Bottom Line
Cantor’s core message is that AI safety should be addressed through liability, competition policy, and basic regulatory guardrails—not through an antitrust exemption that could let a handful of companies coordinate like a cartel. He argues the U.S. should resist both AI hype and national-security panic, and instead force the industry to innovate within clear rules that protect the public.