Latest / Minds, Bodies, and Terawatts / Paying Subscribers Sue AI Leaders Over Coordinated Slowdown

Transcript

Paying Subscribers Sue AI Leaders Over Coordinated Slowdown · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts — tracking the collision of AI, robotics, and energy. Quartz reported today, September 20th, that paying subscribers have filed a federal antitrust lawsuit against OpenAI, Anthropic, Google, and SpaceXAI.
• 0:12
  [Guest] The charge is collusion, but not to raise prices. The plaintiffs allege that when these labs publicly agreed to slow down their AI models for safety, they formed an illegal cartel to throttle the products their customers are paying for.
• 0:26
  [Host] We need to look at exactly what triggered this. Last week, Dario Amodei from Anthropic called for a coordinated industry slowdown on advanced models. Sam Altman from OpenAI publicly agreed within a day, and Elon Musk signed on too. It was universally covered as a rare moment of responsibility—the people building the frontier finally tapping the brakes.
• 0:46
  [Guest] Right, an ethical pledge to not rush blindly into systems that write their own code or act without oversight. But if you are a software engineer paying twenty dollars a month for ChatGPT Plus, or a startup relying on Claude's API to run your backend, you didn't sign a safety pledge. You are paying for the most capable model the market can produce.
• 1:08
  [Host] And that is exactly the argument in the lawsuit, filed in the Northern District of California. The plaintiffs say that when the four biggest players in a market agree to hold back their best technology, that is textbook anti-competitive behavior. It restricts the supply of intelligence.
• 1:24
  [Guest] It is a genuinely brilliant legal trap. Antitrust law does not have a carve-out for saving the world. It only asks if competitors are coordinating to restrict supply, fix prices, or degrade product quality. By agreeing to pace themselves in public, the labs handed their own paying subscribers a perfect argument: you are colluding to make my subscription less valuable than it would be in a free market.
• 1:48
  [Host] This gets very real for the listener. If you rely on these tools, the difference between today's model and the one they are holding back is your personal productivity. If SpaceXAI and Google agree not to ship their smartest coding agents this quarter, an independent developer loses hours of work every week. The plaintiffs are arguing that they are suffering concrete economic damage because the labs decided to play referee.
• 2:12
  [Guest] And think about the scale. OpenAI revenue passed forty billion dollars a year last month. Anthropic booked over eleven billion dollars in a single quarter. The labs thought they were stepping up to solve a governance gap. They forgot they are also massive consumer software businesses. When you have a billion weekly users, and millions are paying you directly, any coordination with your rivals looks like market manipulation. You can't just call a truce with your competitors when you're charging users for a race.
• 2:42
  [Host] Which brings us to the core problem with how we govern these systems. We watched the industry finally agree to a safety slowdown, and the immediate result is a federal lawsuit trying to force their foot back onto the gas pedal.
• 2:56
  [Guest] Because the structural incentives haven't changed. This is the danger of self-regulation in a market driven by extreme network effects and vendor lock-in. These platforms aren't just selling software; they are building memory networks. They accumulate intelligence about you—your coding style, your projects, your thought patterns. The longer you use a specific model, the harder it is to leave.
• 3:17
  [Host] So when the labs agree to slow down, the user can't just easily jump to an open-source alternative. Their entire digital context is hostage to the platform.
• 3:26
  [Guest] Right. They have built an intelligence moat. When those dominant platforms try to govern themselves informally, the law looks at them and sees an oligopoly protecting its position under the guise of safety.
• 3:38
  [Host] The plaintiffs aren't asking for a seat at the safety table. They want the court to declare the safety table itself illegal.
• 3:45
  [Guest] Exactly. An informal agreement between CEOs is incredibly fragile. It has no independent verification, no enforcement, and as of today, it carries massive legal liability. If you want a real pause button, you need pre-negotiated protocols, independent inspectors inside data centers, and liability triggers the government actually recognizes. You need real infrastructure.
• 4:03
  [Host] But the government just told them to self-police! The White House and Beijing spent last week agreeing the labs should handle their own oversight, because nobody in power wants to build a new regulatory agency.
• 4:16
  [Guest] Which leaves the labs completely trapped. The government refuses to regulate them, so they try to regulate each other. But the moment they do, their own customers sue them under antitrust laws designed for the railroad and oil era. You cannot bolt governance onto a compounding system after the fact. We are trying to supervise the most consequential capability humans have ever built using legal institutions from 1890, and the system is breaking.
• 4:41
  [Host] So the labs are caught in a vise between the safety alarm they just pulled, and the subscribers demanding they keep driving.
• 4:49
  [Guest] For more on the concepts discussed, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.