🔍 Read the full analysis: Who Answers When An AI Agent Makes Decisions For You? on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Thorsten Meyer’s ebook Many Hands, published in October 2026, examines responsibility when AI agents buy, pay or accept terms for people and organizations. It argues that legal liability, organizational responsibility and moral blame can fall on different parties; how those questions resolve in specific cases depends on the facts, contracts and applicable law.
Thorsten Meyer published Many Hands: Who Answers When the Agent Acts in Your Name? as an ebook in October 2026, examining who is responsible when an AI agent makes purchases, accepts terms or otherwise acts for a person or company. The book separates the question into who pays for a loss, who must explain and fix it, and who should be held morally responsible—questions that may not have the same answer.
Meyer introduces the issue through a composite scenario, not a reported real-world incident. In it, a purchasing agent accepts a steel supplier’s offer at a price seven percent lower, committing a company to buy 120 tonnes over 12 months for about €370,000. The owner had separately agreed to divide the next year’s volume with another supplier, but that agreement was not available to the agent. Meyer writes that the agent followed its rules and recorded its actions, even though the decision conflicted with what the owner wanted.
The example illustrates the book’s central distinction: an agent can follow its programmed rules while producing an unwanted result. Meyer argues that deciding who is legally liable differs from determining who inside an organization has to investigate and repair the problem, and from deciding who bears moral responsibility. The book does not present the steel purchase as a verified event; it uses the fictionalized case to frame those questions.
The ebook’s key recommendations include assigning a named person to each agent, limiting its permissions to the task, keeping records of who made decisions, and giving affected people a way to raise concerns. Meyer also argues that records should support learning rather than simply punishment. These are proposals in the book, not a description of a new legal requirement.
Many Hands · the core idea
One question, three answers
"Am I liable if the agent does it in my name?" sounds like one question. The book splits it into three, because agent systems pull the answers apart.
Decided by courts, contracts and insurers, on their own timetable.
In Theresa's caseHer company will almost certainly pay for the steel, whatever it later recovers from others.
Decided inside the organization: who must say what happened and what will change.
In Theresa's caseTheresa explains it to the second supplier, her staff and perhaps her bank.
Who owes the people affected an acknowledgment. Often answered last, if at all.
In Theresa's caseStill open when she reaches the motorway.
The hands behind one overnight order
Introduction · a composite scene from the book. None of them saw the acceptance before it went out.
- Model makertrained the language model
- Vendorbuilt the agent on top of it
- Consultantwrote its rules with her staff
- Supplier's portalaccepted the declaration at 2:14 a.m.
- Her companyswitched it on; Theresa signed the rules
“The moral crumple zone protects the integrity of the technological system, at the expense of the nearest human operator.”Madeleine Clare Elish (2019), quoted in Chapter 1
Source: Thorsten Meyer, Many Hands: Who Answers When the Agent Acts in Your Name? (ebook, October 2026; law as of 16 October 2026), Introduction and Chapter 1. Theresa is a composite built from common situations, labeled as such in the book.
Powered by Thorsten Meyer AI
When Agent Decisions Create Business Losses
The accountability question matters as agents move from generating text to taking actions such as buying, paying and accepting terms. A mistaken answer may be corrected; an accepted contract or payment can create a financial commitment. Organizations using agents may need to determine not only what the system did, but who supplied its instructions, what information it had, and who had authority to stop it.
Meyer argues that responsibility can be dispersed across the user, employer, software provider and people who design or oversee the system. That makes it harder to identify who should absorb a loss or change procedures. His framework is a way to examine those separate questions, rather than a conclusion that any particular party is always responsible. Legal outcomes remain dependent on the facts, contracts and governing law.
The distinction also matters to people affected by an automated decision. A usable process for review and correction could give them a route to challenge an action, even where the internal record shows that the agent followed its instructions. The book advocates such channels, but the supplied material does not establish that they are already standard practice or legally required in every setting.
Popular right now
Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
Shark Rocket Pet Plus Corded Stick Vacuum, Bordeaux, HV322
As an affiliate, we earn on qualifying purchases.
From Many Hands to Agent Systems
The book takes its title from political philosopher Dennis F. Thompson’s “problem of many hands,” a term Meyer traces to 1980. It describes the difficulty of assigning responsibility when many people contribute to an outcome. Meyer’s argument is that AI agents add more participants and distance between decision-makers, while acting faster than conventional processes.
To show how responsibility can settle unevenly, Meyer discusses the 2018 fatal crash involving a self-driving test vehicle in Tempe, Arizona. He cites the US National Transportation Safety Board’s account that the system detected pedestrian Elaine Herzberg 5.6 seconds before impact, did not settle on what she was, and was designed to withhold emergency braking for one second. The board identified factors involving the operator and company, as well as the state and pedestrian. Meyer notes that the criminal case reached the human operator closest to the event, while asking how responsibility should be understood across the wider system.
Meyer also refers to the Dutch childcare benefits scandal, in which tens of thousands of families were wrongly required to repay benefits. He describes how responsibility can be spread across risk models, officials and other institutions. These examples predate the ebook and serve as background for its argument; they are not cases about AI agents making commercial purchases.
“protects the integrity of the technological system, at the expense of the nearest human operator.”
— Anthropologist Madeleine Clare Elish, as quoted by Meyer
Liability Depends on the Specific Case
The ebook raises questions rather than resolving how a court would treat every agent decision. Who pays for a particular loss can depend on applicable law, the parties’ contracts, the agent’s authority and the facts surrounding its use. The source material does not provide a legal ruling on the fictional steel commitment or establish that one party would necessarily bear its cost.
Meyer says European product-liability rules cover software and certain forms of injury, private-property damage and private-data harm, while excluding pure financial loss such as a bad contract or incorrect payment. That is his description of the rules. The supplied material does not include the relevant legal text or establish how the rules apply across different jurisdictions, products or claims.
The source also gives no detailed account of the book’s recommendations being adopted by a company or regulator. The status of any agent’s authority, safeguards and review process in a real dispute would need to be established separately.
The Debate Moves to Real Deployments
Many Hands was released as an ebook in October 2026. Meyer lists availability through Apple Books, Barnes & Noble, Kobo, Thalia, Bookshop.org, Smashwords, Fable and Vivlio, and describes an accompanying video that presents the book’s argument in five minutes.
The practical test for the book’s framework will come in how organizations set up and oversee agents that can take consequential actions. Questions to watch include whether each agent has a clearly identified owner, whether its permissions match its task, and whether people affected by its decisions can seek review. The source material does not identify a specific regulation, court decision or company rollout as the next milestone.
For readers assessing a real incident, the unanswered questions remain concrete: what the agent was authorized to do, what information it received, who could intervene, and how the resulting loss will be handled. Those facts, rather than the presence of an AI system alone, will shape who has to answer.
Key Questions
What is Many Hands about?
Thorsten Meyer’s October 2026 ebook examines accountability when AI agents act for people or organizations. It separates legal responsibility, organizational duties to explain and fix problems, and moral responsibility.
Is the steel purchasing story a real incident?
No. Meyer identifies Theresa and the steel purchase as a composite scenario built from common situations. It illustrates the book’s argument and is not presented as a verified transaction.
Does the book say who pays when an agent makes a mistake?
It does not give one universal answer. Meyer says legal liability depends on courts and contracts, while an organization’s responsibility to investigate and the moral question of who should have acted differently are separate issues.
What safeguards does Meyer propose?
His recommendations include naming a person responsible for each agent, limiting permissions to the task, recording who made decisions and giving affected people a way to raise concerns. These are proposals in the book, not universal legal requirements established by the source.
Source: ThorstenMeyerAI.com
