🔍 Read the full analysis: Who Is Responsible For What An AI Agent Does In Your Name? on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Thorsten Meyer’s ebook Many Hands, released in October 2026, examines responsibility when AI agents take actions such as buying goods or accepting contract terms. It separates who pays under law and contracts, who must explain and fix a failure inside an organization, and who bears moral responsibility; the practical and legal answers can differ.
Thorsten Meyer has released Many Hands: Who Answers When the Agent Acts in Your Name?, an ebook about who is responsible when an AI agent buys, pays or accepts terms for a person or organization. The book argues that the question has three separate answers: who pays, who must explain and repair the outcome, and who should have acted differently.
Meyer opens with a composite scenario: a purchasing agent accepts a steel supplier’s terms at 2:14 a.m., securing a seven percent price reduction while committing its company to buy 120 tonnes over 12 months, with no cancellation right. The commitment is valued at about €370,000. The owner, Theresa, had agreed the previous day to divide next year’s volume with another supplier, but the agent had not been told. Meyer says the example is constructed from common situations, not a reported real-world transaction.
The scenario illustrates the difference between an agent following its instructions and an organization getting the outcome it wanted. Meyer writes that, under the agent’s rules, it “had made no mistake.” In the scene, the purchasing lead says: “Nothing in it was wrong, and nothing in it was what she wanted.” The book asks how authority, missing information and human review should be handled when an automated system can make commitments in a company’s name.
Meyer divides responsibility into legal, organizational and moral questions. Courts and contracts determine who bears financial liability; organizations decide who must explain and fix a failure; moral responsibility concerns who should have acted differently. The book’s stated safeguards include a named person responsible for each agent, permissions limited to the task, records of decisions, a way for affected people to seek help, and use of those records to learn rather than punish.
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 Bind Businesses
The issue matters because AI agents are increasingly framed not only as tools that generate answers but as systems that can book, buy, pay and accept terms. When those actions create a contract or financial loss, an organization may have to respond even if no individual intended the precise result. The steel example shows how a lower unit price can come with a larger commitment that conflicts with a separate agreement.
Responsibility can also be distributed across the people who set up an agent, approve its access, supply its information and oversee its operation. Meyer’s argument is that this can make it harder to identify who should answer for a bad outcome, while the person closest to the final action may be easiest to blame. The distinction between technical compliance and organizational intent is particularly relevant to companies deciding what decisions agents may make without approval.
The ebook is an author’s analysis, not a court ruling or a new legal standard. Its examples and recommendations raise practical questions for employers and developers, but they do not settle liability in a particular dispute. Anyone affected by an agent’s action would still need to consider the relevant facts, contracts and applicable law.
Products Worth Considering
Trezor Safe 5 Crypto Hardware Wallet with Color Touchscreen, Black Graphite
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
Ledger Stax Crypto Wallet with Touchscreen
As an affiliate, we earn on qualifying purchases.
From Many Hands to AI Agents
The book’s title draws on political philosopher Dennis F. Thompson, who described the “problem of many hands” in 1980: when many people contribute to an outcome, it can be difficult to assign responsibility. Meyer argues that agents add more participants and distance between decision-makers, while acting faster. He writes that an AI agent itself can neither pay nor apologize, leaving people and organizations to handle the consequences.
Meyer connects the question to earlier cases involving automated systems. One is the 2018 self-driving test-car crash in Tempe, Arizona, in which Elaine Herzberg was killed. Citing the US National Transportation Safety Board, he says the car detected her 5.6 seconds before impact, did not settle on what she was, and by design delayed emergency braking for one second. The case is used to examine how responsibility can fall on a nearby human operator even when system design and organizational oversight are also at issue.
Another case is the Dutch childcare benefits scandal, where families were wrongly required to repay benefits. Meyer says more than 44,000 people had been recognized as affected in a government report from September 2026. He also discusses European product-liability rules, arguing that their coverage of software-related harm does not extend to pure financial losses such as an incorrect payment or bad contract. These points are presented as part of the book’s analysis; the exact legal outcome in any agent dispute depends on the applicable rules and facts.
“An agent “can neither pay nor apologize.””
— Thorsten Meyer, on the responsibilities an AI agent cannot take on
Liability Depends on the Specific Case
The ebook does not establish how a court would decide a future dispute involving an AI agent. Legal responsibility will depend on the jurisdiction, the contract, the agent’s permissions and the actions of the people involved. Meyer’s account of European product-liability rules is his interpretation of the scope of those rules, not a ruling that all agent-related financial losses fall outside legal remedies.
The opening steel purchase is explicitly a composite scenario, so it does not show that a particular company incurred a €370,000 loss. The source material also does not identify a real transaction behind it. It remains unclear how often agents currently accept binding commercial terms without human approval, or what proportion of such actions result in disputes.
Nor does the book announcement provide a detailed account of how different vendors’ contracts allocate losses. Meyer says some model-provider terms cap liability at about a year of fees, but the specific contracts and conditions are not supplied here. Readers should not treat that figure as a universal limit across providers.
Putting Accountability Into Practice
The book’s proposed next step is operational rather than a new legal rule: organizations can identify a person responsible for each agent, give it only the permissions needed for its assigned task, and retain records showing who authorized what and when. Meyer also argues that affected people need a clear route to raise a problem and that records should support learning, not simply punishment.
For companies using agents in purchasing or other high-impact work, those proposals point toward reviewing approval thresholds, contract authority and escalation procedures. Whether such measures reduce disputes, and how regulators or courts may treat them, remains to be seen. Meyer’s ebook, published in October 2026, sets out the questions; it does not resolve them for every organization or legal system.
Key Questions
What is Many Hands about?
Thorsten Meyer’s ebook examines who is responsible when an AI agent acts on someone’s behalf. It separates who pays, who must explain and fix the result, and who bears moral responsibility.
Is the steel purchase in the book a real case?
No. Meyer describes Theresa and the €370,000 steel commitment as a composite scenario built from common situations, not as a documented real transaction.
Does the book say AI agents are legally liable?
No. The book’s framing is that an agent cannot pay or apologize; people, companies, contracts and courts determine how consequences are handled. The legal answer depends on the circumstances and applicable law.
What safeguards does Meyer propose?
He proposes naming a person responsible for each agent, limiting its permissions to its task, keeping decision records, providing a route for affected people to raise concerns, and using records to improve processes.
When was the ebook released?
Many Hands: Who Answers When the Agent Acts in Your Name? was published as an ebook in October 2026.
Source: ThorstenMeyerAI.com
