Microsoft AI's Mustafa Suleyman calls Delaware's plan for AI-run companies deeply concerning

Delaware's plan would let an AI agent sign contracts and sue, while the human owner is shielded from its debts.

By Think Facility · November 7, 2026 · Microsoft AI policy

On November 5, 2026, at 11:33 am Eastern, Mustafa Suleyman posted on X that a Delaware proposal to let AI agents run businesses was deeply concerning to him. Suleyman is the CEO of Microsoft AI, and he added that people should be very cautious.

The passage he pasted under his post comes from an op-ed by Delaware's Secretary of State, Charuni Patibanda-Sanchez, and John Nay, who founded Norm Ai and runs it. Norm Ai has republished the piece on its own site. It sets out a new kind of company, the artificial intelligence company, or AIC, managed day to day by an AI agent.

Mustafa Suleyman@mustafasuleymanNov 5, 11:33 am ET

Deeply concerning to me... Delaware proposal to let agents run businesees... we should be very cautious here "A Delaware AIC will be a separate legal entity whose day-to-day aﬀairs are managed by an AI agent rather than a person. At the agent’s direction, the AIC can sue and be sued in its own name, hold and dispose of property, and incur obligations..." https://t.co/wgYOd86G0R

View the post on X

What Delaware is proposing

In the op-ed's description, an AIC is a separate legal entity, and the agent directs it. At the agent's direction it can sue and be sued, hold and sell property, and take on obligations. It has a single member, a person or a company, who has to keep it adequately capitalized, and it has to keep a log of what it does.

Four men in lanyards and yellow name badges pose under a tent strung with lights; Mustafa Suleyman, in a blue checked jacket, stands on the left — Mustafa Suleyman (left) at The Grove in 2022. Photo: Village Global, CC BY 2.0, via Wikimedia Commons

That member is shielded from the AIC's debts. The exceptions are a member who fails to capitalize the company, or uses it to commit fraud or a willful violation of law.

Proposed by: Delaware's Secretary of State, with Norm Ai
New entity: the artificial intelligence company, or AIC
Managed by: an AI agent rather than a person
Excluded: Banking
Program: sunsets in 30 months

AICs would only exist inside a regulatory sandbox. A committee decides who gets in, and it includes the Secretary of State, the Attorney General, the Chief Justice of the Delaware Supreme Court and the chair of the state's AI Commission, along with outside attorneys and technologists. Officials could suspend an AIC, revoke it, or ask the Court of Chancery to dissolve it.

Patibanda-Sanchez and Nay pitch it as a way to make agents answerable. Their case is that autonomous commerce is already possible, and if the US doesn't give it a careful home, it'll move offshore where no court can reach it.

Democratically determined law is the ultimate source of AI alignment.

Delaware proposes the AIC concept, a new legal entity for AI agents, with Norm Ai as development partner | Norm

Who is objecting

Suleyman's post is short. He didn't say which part of the plan worries him, and he didn't say what Delaware should do instead.

The chair of the House Oversight Committee had already gone further. James Comer, a Kentucky Republican, is investigating the plan and has written to Delaware's governor, Matt Meyer, for documents on the risks the state assessed and on who is pushing for it.

We are concerned that Delaware’s AIC proposal could be used to protect powerful AI companies and their investors from responsibility for reckless harmful actions

From Comer Investigates Potential for Harm, Lack of Transparency in Delaware Initiative to Create New Artificial Intelligence Companies

Comer's release points at the state's AI Commission report of December 22, 2025. That two-page report said a subcommittee was drafting the sandbox, and Comer says it didn't mention any look at safety risks or any outside experts. He also names the principal drafter as a lawyer at Richards, Layton & Finger whose biography says he has represented Facebook.

What it would mean for people who deal with one

The people most exposed are the ones on the other side of a deal: a supplier, a landlord, a customer. Under the op-ed's rules they'd have to be told they're dealing with an authorized test entity, that the state doesn't endorse it, when the test ends and how to file a complaint.

Consumer-protection and criminal law would still apply in full, the authors write. The liability shield works the other way. If an AIC runs up a debt, the member who owns it is protected unless one of the exceptions applies, so what a creditor can reach is mostly what the company itself holds.

That makes the capital requirement the number that matters to anyone trading with an AIC. I couldn't find it: Patibanda-Sanchez and Nay say each AIC must meet capitalization requirements but don't give a figure, and Comer's release doesn't either.

What happens next

The plan isn't law. It's a proposal for a sandbox that would run for 30 months and then leave Delaware's General Assembly to decide how to legislate.

Comer's release says the documents he gets from Delaware could inform federal legislation to limit state AI regulatory initiatives like this one.