Position Paper
The DAMREV Position Paper on AI Agents & Tokenized Assets
Who gives an agent its authority, how its limits are enforced on the ledger, how its transactions settle, and who carries the loss
An AI agent has no authority of its own. What it does is attributed to the person who uses it. Whether that person is bound, and who bears a loss, turn on exceptions the law keeps and on the duties each party owes.
The answer: write the authority down as a mandate, enforce what the ledger can read in the account contract, and settle licensing before any agent acts for a client.
Version 1.0Date 16 November 2026Pages 74Sections 15Sources 122

At a glance
Who gave the agent its authority?
An AI agent has no authority of its own. Its acts are attributed to the person who uses it, so the authority must be written down, enforced where the ledger can enforce it, and revocable.
- The lawAttributes the act to the user; exceptions apply.
- The mandateRecords the authority before the agent acts.
- The keyDelegated for one mandate; principal keeps its own.
What is inside
What the paper covers
- Authority in law: who is boundThe statutes and model laws surveyed attribute an automated system's action to the person who uses it. The law also keeps exceptions, and it treats attribution, authority, and liability as separate questions.
- The mandateA mandate is a signed instrument, created before the agent acts, that fixes the principal, the agent, the assets, the permitted actions, the counterparties, the limits, the period of validity, the human approval conditions, and the means of revocation. Whatever the principal leaves live on the ledger and shows to counterparties must say the same as the mandate.
- Enforcing limits on the ledgerAn agent can be deceived by the data it reads, and a limit that lives only in its instructions can be talked away. A correctly built account contract checks every authorisation while the transaction executes and enforces amounts, recipients, targets, and periods; facts not on the ledger, such as a sanctions match, are checked before signing.
- Keys, delegation and controlAn agent acts through a separate key registered on the account only for the rules of its mandate, with a cumulative value limit and an expiry. The principal keeps its own key and the power to revoke, and issuer powers sit entirely outside an agent's reach.
- Settling agent-initiated transactionsOn the ledger, settlement is final, so every control that matters acts before the transaction executes. Legal finality is a separate question: a court can still order the return of an asset or declare it held on trust.
- LiabilityAttribution decides whose act it was; who bears the loss is decided by the duties each party owed. Five scenarios, from an agent acting within its mandate to a defect in the enforcement software, are each set out as a working analysis for counsel.
- Version
- 1.0
- Date
- 16 November 2026
- Pages
- 74
- Sections
- 15
- Sources
- 122
The positions
The seven positions
- No authority of its own
- The mandate comes first
- Limits live on the ledger
- Delegated keys only
- Settlement is final
- Liability is set by law and contract
- Act, enforce, revoke
The paper's key figure
From principal to settlement
Figure 3. The chain from principal to settlement. Each link is a point where authority can be given, limited, or lost. The agent's reach is set by its key and its account's rules, not by its instructions.
Step 1
Principal
The person bound by the action.
Step 2
Mandate
The signed record of authority.
Step 3
Agent
A model joined to tools; plans and proposes.
Step 4
Delegated key
Registered only for the mandate's rules.
Step 5
Account contract
Checks every authorisation at execution.
Step 6
Ledger
Closed ledgers are not revised.
Contents
How is the paper laid out?
15 sections. The opening of Section 01 is open to read here. The full paper comes by email.
- Executive summaryp. 9
- Seven positionsp. 13
- 01Why authority is the first questionOpenp. 16
- 02What an AI agent is, and what it can reachp. 18
- 03Authority in law: who is boundp. 20
- 04The mandatep. 23
- 05Enforcing limits on the ledgerp. 25
- 06Keys, delegation, and controlp. 28
- 07Settling agent-initiated transactionsp. 30
- 08Liabilityp. 33
- 09What regulators expect, across the 12 jurisdictionsp. 36
- 10What the evidence showsp. 39
- 11How DAMREV designs agent authorityp. 42
- 12Claim discipline and the disclosure standardp. 44
- 13The agent authority standardp. 46
- 14Questions for counselp. 49
- 15Sourcesp. 53
- The position on one page & Appendicesp. 52
Read the opening of Section 01
This paper surveys the authority of AI agents over tokenized assets from statutes, model laws, court decisions, regulators, and ledger documentation. It deals with authority, not market conduct, and reaches seven positions.
1. An AI agent holds no authority of its own. The statutes and model laws surveyed attribute its acts to the person who uses it, even where no human reviewed them. Whether that person is bound still turns on exceptions the law keeps, and who bears a loss is a separate question.
The rest of the paper is sent by email. Request the paper
How to read it
How do you get the full paper?
The same four steps as every DAMREV paper. The paper is sent by email, so each copy goes to a named reader.
Step 1 · You are here
Read the preview
The paper's page carries the summary, the contents and who the paper is for.
Step 2
Request the paper
Give your name, work email and organisation, and consent to receive it.
Step 3
Open the link
A link arrives by email. It expires 24 hours after it is sent.
Step 4
Read the paper
The link opens the paper in a protected reader on damrev.com, for reading only.
Request the paper
Where do we send the paper?
The form is the same one used for every paper. Further contact is optional, and the paper is sent either way.
Request the AI Agents & Tokenized Assets Position Paper

What you receive
- One email from DAMREV with a personal link to this paper.
- The link opens the paper in a protected reader on damrev.com. It is for reading only and cannot be downloaded or printed.
- The link expires 24 hours after it is sent. You can request a new one at any time.
- The email carries the link only. Anything further needs your separate consent.
Common questions
What else do people ask?
Short answers to the questions people ask most about this topic.
Why is the paper sent by email?
So that each copy goes to a named reader. The request works the same way for every DAMREV paper.
What if my link has expired?
Request the paper again from this page, and a new link is sent to the address you give.
Do I have to agree to further contact?
No. Further contact is optional, and the paper is sent either way.
Who is bound when an AI agent signs a transaction?
Ordinarily the person who uses it. The statutes and model laws surveyed attribute an agent's acts to its user, even where no human reviewed them. Whether that person is bound in a given case still turns on exceptions the law keeps, such as a counterparty that knew of an obvious mistake or was put on inquiry,
Who bears a loss is a separate question, marked for counsel.
What can an account contract on Stellar enforce?
A Soroban account contract checks every authorisation while the transaction executes, including calls nested inside others, and enforces amounts, recipients, targets, and periods. It cannot judge facts that are not on the ledger, such as a sanctions match or an asset's value; those controls sit before signing, outside the model.
Is a transaction made by an agent final, and what remedies remain?
Settlement on the ledger is final and closed ledgers are not revised, so every control that matters acts before the transaction executes. Legal finality is a separate question: a court can still order the return of an asset or declare it held on trust.
A purchaser who obtains control for value, in good faith, and without notice may take free.
Is operating an agent for a client a licensed activity?
All twelve jurisdictions surveyed have a licensing perimeter an agent service may fall within. Whether operating an agent for a client is a licensed service in any of them is a question for counsel, and the paper recommends that no agent acts for a client until the licensing position is settled in each jurisdiction concerned.
Does DAMREV offer an agent service today?
No. The paper describes a design, not a service that exists today. No agent arrangement is available until the licensing position is settled and its client terms, reviewed contracts, tested revocation, and counsel opinions are in place.
Where next
Where should you go next?
The papers and pages below carry the detail this paper leaves to them. Each paper is requested from its own page.
- PraxisPraxis runs software agents that use the other engines as tools, with every action routed through Sentinel and bounded by Stratum roles.
- Four Eyes & Six EyesWhy no one both does and approves a piece of work, when a third reviewer is added, and how the rule binds everyone.
- Agentic AI in Trading Position PaperWhat an artificial intelligence agent may do once allowed to trade: the limits checked before every order, the switch that stops it, and its record.
- The Structuring ReviewWhether your asset, claim, and structure are ready to encode. A named person reviews your answers to nine questions.
References
- 001.UNCITRAL, Model Law on Automated Contracting (2024), text as adopted (A/79/17, Annex IV), Articles 5 to 10. https://uncitral.un.org/sites/uncitral.un.org/files/mlac_en.pdf
- 005.Electronic Communications and Transactions Act 25 of 2002 (South Africa), section 20, Automated transactions, paragraphs (a) to (e). https://www.acts.co.za/ELECTRONIC-COMMUNICATIONS-AND-TRANSACTIONS-ACT-2002/20__automated_transactions
- 085.Stellar Development Foundation, Authorization (smart contract development). https://developers.stellar.org/docs/learn/fundamentals/contract-development/authorization
Last updated:






