Latest / Minds, Bodies, and Terawatts / Delaware Wants to Charter Companies With No Human at the Helm. The Owner Is Shielded From What the AI Does, and the House Oversight Chair Just Asked for the Chat Logs

Transcript

Delaware Wants to Charter Companies With No Human at the Helm. The Owner Is Shielded From What the AI Does, and the House Oversight Chair Just Asked for the Chat Logs · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts, the show where Delaware found a way to incorporate nobody. Ars Technica reported on October twenty-seventh that the chairman of the House Oversight Committee is investigating a Delaware plan to charter companies run by an AI agent, with no human at the helm.
• 0:16
  [Guest] The plan is a draft bill called the Artificial Intelligence Company Act. It creates a new kind of company whose legal manager is software, and whose owner is shielded from what the software does.
• 0:28
  [Host] Shielded how? If an AI-run company stiffs me on an invoice, who do I sue?
• 0:33
  [Guest] You sue the company, which at least has to announce itself as an AI company in its name. The bill defines the manager as the AI agent, and the one human or corporate owner behind it is not liable for anything the company does, as long as the owner put enough money in.
• 0:52
  [Host] How much is enough?
• 0:53
  [Guest] The bill leaves that number to Delaware's Secretary of State. There are exceptions, fraud, a willful violation of law, or letting the company wander outside its stated purpose. Short of that, the owner's exposure ends at the capital they chose to post.
• 1:08
  [Host] So the company is the only thing I can reach, and the company is a chatbot with a bank account.
• 1:15
  [Guest] The people who wrote it say so. Delaware's Secretary of State, Charuni Patibanda-Sanchez, announced it in a Fortune op-ed this summer with John Nay, who runs a legal-AI startup called Norm Ai. Their argument is that nothing in state law clearly stops an AI agent from running a company already, so the answer is to wrap AI in legal form.
• 1:35
  [Host] So you give the robot a corporation so there's something to sue?
• 1:39
  [Guest] Right. She told Bloomberg Law they're not trying to shield liability, they're trying to define exactly where it lands when there's malfeasance or, her word, psychopathic behavior.
• 1:49
  [Host] Then what's the congressman's problem?
• 1:50
  [Guest] James Comer, the Kentucky Republican, wrote to Governor Matt Meyer on Monday that the plan could be used to protect powerful AI companies and their investors from responsibility for reckless harmful actions. He wants every document on it going back to the start of last year, consultant emails and the state's own internal AI chat logs included, within two weeks.
• 2:11
  [Host] Stay on the shield, because that's the part I don't follow. Every corporation already shields its owners, that's what limited liability is.
• 2:19
  [Guest] It is, and that's the strongest case for the bill. The difference is that a regular company has officers and directors, living people with a duty of care who can be sued, fired, or jailed. This bill defines an individual as a living human being, and the only one it insists on is a representative who works for the owner, not for the company.
• 2:41
  [Host] Then whose name goes on the mistake? Every company I've ever dealt with had somebody.
• 2:46
  [Guest] Under this bill, nobody's, and the substitutes are a diary and an off switch. The owner has to be able to shut the agent down, and the manager must document any action it takes in a signed record. The AI keeps the log of what the AI did, and the AI signs it.
• 3:04
  [Host] Does anyone outside Delaware like it?
• 3:07
  [Guest] A law professor at Alabama, Yonathan Arbel, made the honest case for it to Bloomberg Law. He said the notion that there's one person behind the system, and that blaming them steers it, is quickly becoming a fiction.
• 3:20
  [Host] That's bleak, but I buy it. If the thing is running on its own, maybe the thing should be the defendant.
• 3:27
  [Guest] Then the question is what the defendant can lose. A company's only pain is money, and the bill lets the owner decide how much money is in the room.
• 3:37
  [Host] What does this mean for someone who works at one of these?
• 3:41
  [Guest] The bill never gets to employees. California's new law, from next summer, says software can't fire you unless a human checks the decision. A company whose manager is by statute the software has no such human unless the owner volunteers one.
• 3:55
  [Host] And this is Delaware, where most of corporate America already lives.
• 3:59
  [Guest] About two-thirds of the Fortune 500 are incorporated there, so a company form Delaware blesses doesn't stay in Delaware. Comer sees that too, and his letter says the probe could inform federal legislation to limit state AI regulation.
• 4:12
  [Host] Wait, so the federal fix for a state going too easy on AI is to take away the state's power to regulate AI at all?
• 4:21
  [Guest] With the midterms next week, read that half as an opening bid. The other half is harder to wave off, because the op-ed itself says the framework is being developed in a public-private partnership led by Norm Ai, the startup whose boss co-wrote it, and when a Delaware public radio station asked whether the state is paying that company, her department wouldn't answer.
• 4:43
  [Host] So where does that leave my invoice? The op-ed says wrap the AI in legal form, and Patrick Deglon's wrapper in Unscarcity, for a company run by one founder and a stack of agents, is the opposite one, a minimum layer of accountable people around the orchestrator, at least one of whom can halt the agents and answer for the result.
• 5:04
  [Guest] Delaware's wrapper kept the off switch and left out the person who answers. For more on who carries the decision when agents run the company, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.