Part five · First principles ·
Take On More Liability.
For a century, firms have spent fortunes limiting what they answer for. In the age of super intelligence that is exactly backwards. Customers, vendors and partners will all want to deal with the entity that assumes the liability, and that firm will do very well.
By John Rector · about 8 minutes
Liability is gravity
Two firms, same delegates. Both firms disclaim what they can. Customers, vendors and partners split between them.
- Customers
- Vendors
- Partners
- A firm that limits its liability
Liability is gravity. Be the center.
The old playbook
Disclaim, cap, carve out, arbitrate. Sell the work and keep the risk at arm’s length.
The new one
Assume it. Price it, back it, and stand behind the outcome, because the outcome is the only thing left that is scarce.
Monday, 9:05 a.m.A restaurant owner reads two contracts for the same AI service. Both firms run excellent delegates. Both will do the same work. She reads the first contract to page eleven, then stops.
Services provided as is. Provider makes no warranty that outputs are accurate. Total liability limited to fees paid in the prior twelve months.
If our work causes you a loss, we pay it.
The firm’s whole advantage, given away
In part four I argued that most customers will keep choosing a firm over an individual or a small team for one reason: the firm can be insured. It can carry a limit. It can answer for things. That is the firm’s whole advantage in the age of super intelligence.
Then look at how most firms spend their legal budget. They write limitation-of-liability clauses, warranty disclaimers, caps tied to fees, carve-outs, arbitration requirements, and the two most expensive words in commerce: as is. All of it shrinks the one thing customers are coming to the firm to buy.
That playbook made sense when the work was scarce. The firm’s value was the work, and the risk was a cost to keep small. Delegates have turned that around. The work is heading toward the cost of a sentence. What remains scarce is someone who will stand behind it.
A firm that disclaims the outcome is selling the cheapest thing in the market and refusing to sell the only scarce one.
The firms that went first
Some firms saw this early. In 2015 Volvo’s chief executive, Håkan Samuelsson, said the company would accept full liability whenever one of its cars was in autonomous mode. His reasoning was a single line: “We are the suppliers of this technology and we are liable for everything the car is doing in autonomous mode.”
Mercedes-Benz made the promise real with Drive Pilot, a Level 3 system. When it is engaged, Mercedes accepts legal responsibility for what the car does. The technical difference between Level 2 and Level 3 driving is mostly a difference in who is liable: the driver, or the machine’s maker.
In 2023 Microsoft did the same for AI output. Under its Copilot Copyright Commitment, it promised to defend commercial customers who are sued for copyright infringement over Copilot output, and to pay any adverse judgments. Its explanation is the thesis of this whole piece: if customers’ use “creates legal issues, we should make this our problem rather than our customers’ problem.”
Notice what each promise came with. Mercedes takes liability only on mapped roads and in specific conditions. Microsoft requires customers to use its guardrails. That is not hedging. It is the discipline that taking on liability forces. You can only assume what you control, so the firm that assumes liability gets very serious about controlling its delegates.
Why liability is gravity
Every relationship in commerce rests on one question: when something goes wrong, who holds the loss? Whoever holds it becomes the center of mass, and everyone else arranges themselves around it.
Customers want to buy from the entity that holds the loss, because then they don’t have to. Vendors want to supply it, because it can pay and it will keep standing. Partners want to be seen next to it, because their own customers trust it. Lenders, investors and employees follow for the same reason: their exposure ends where the firm’s begins. A firm that pushes liability outward pushes its relationships outward with it.
What it takes to carry more
Assuming liability you cannot pay is not courage. It is a future bankruptcy. So the firms that win this will do three things.
They will build the capacity to pay: balance sheet, insurance, reinsurance, and their own captive insurers where carriers will not go. They will control what they promise: delegates run under real supervision, every action logged, the conditions of use designed the way Mercedes maps its roads. And they will price it openly. The liability a firm carries is its product, and it should be the line on the invoice that customers are happiest to pay.
A council meeting, 2029A town picks a vendor to process building permits with AI. Five bids, five sets of excellent delegates. The winning bid’s first sentence reads: “We are liable for every permit our system processes.” It was not the cheapest bid. Nobody on the council asked about price again.
Five predictions
“We are liable” becomes the headline of the pitch.
It moves from the fine print to the first page. Firms will compete on what they assume, the way they once competed on features.
Liability caps become a competitive disadvantage.
Buyers’ AIs will read every contract and rank vendors by what they will answer for. A cap at fees paid will read like a confession.
The balance sheet becomes the product.
Firms will raise capital specifically to carry more liability, and the chief financial officer’s most important number will be how much the firm can stand behind.
Firms that assume liability will run better delegates.
When a delegate’s errors are the firm’s errors, the firm invests in supervision, testing and records. The quality gap between firms that assume and firms that disclaim will widen every year.
Platforms that assume liability for the work on them become the new giants.
A firm of one will plug into whoever is willing to stand behind her work, which is the backer from part four at the scale of a marketplace.
The thing we were already doing
At Charleston AI we work pay-after: we do the work, the customer receives it, inspects it and uses it, and only then pays. Until they are satisfied, the risk sits on us. And we do it face to face, never over Zoom, because a face is something a customer can hold to account. I described that as a service policy. Writing this series, I see it was the strategy all along. We were taking on liability before I had a word for it.
The claim, and how it fails
- Already true
- Volvo, Mercedes-Benz and Microsoft have each publicly assumed liability for what their automated systems do, and each made it a selling point. Standard insurers are pulling back from AI risk, which leaves the firms that will carry it with the market.
- What has to happen
- Firms must be able to back what they assume with capital, insurance and reinsurance. Laws must let them take the liability on. Even Mercedes has had to work state by state.
- Where I am probably wrong
- Correlated failure. A firm that assumes liability for a delegate used by thousands of customers can be wiped out by one systematic error. If the first big AI liability blowup bankrupts a firm that went first, boards will run back to caps and disclaimers, and this shift will stall for a decade. In that case I was early, not wrong, but early is expensive.
Monday, 9:05 a.m. Again.The restaurant owner puts down Contract A and picks up Contract B. One page, one promise. She signs it before lunch and tells two friends by dinner.
Disclaim the outcome and you are a vendor. Assume it, and you are the firm everyone wants to stand next to.
Background
- Volvo Cars, US urged to establish nationwide Federal guidelines for autonomous driving, October 7, 2015.
- Can’t accept autonomous liability? Get out of the game, says Volvo, Autoblog.
- Mercedes accepts legal liability for Level 3 Drive Pilot system, CarExpert.
- Microsoft, Microsoft announces new Copilot Copyright Commitment for customers, September 7, 2023.
- Mercedes: We’ll be liable for self-driving cars, Kelley Blue Book.
- The series: part one, part two, part three, part four.