For general counsel & risk owners
Your agents can bind the company. You must be able to defend it.
When an agent settles terms with an outside counterparty, the company is on the hook. Subchain holds any material commitment at an approval gate, and records who authorised what for later scrutiny.
What Subchain gives a general counsel.
The approval gate before terms bind
Commitments above a configured value, risk or irreversibility threshold halt and route to the accountable governor. Nothing above your limit binds without an explicit, logged approval.
An immutable record of authorisation
Every proposal, counter, revision, approval and acceptance is written to an immutable log that shows what was agreed and who authorised it.
Evidence for DIFC dispute resolution
The record is assembled as a defensible, evidence-quality account, suited to compliance review, internal audit and DIFC dispute resolution.
Inline policy-as-code checks
Compliance and policy-as-code checks run during negotiation and flag terms that breach configured rules before they reach acceptance, rather than after the fact.
Why an agent commitment is a fiduciary problem.
The commitment binds the company, not the agent
An AI agent has no legal personality and no assets. When it agrees terms with an outside counterparty, the obligation lands on the company that deployed it, and on the officers who answer for that company. That is the exposure a general counsel carries the moment agents begin transacting externally: the organisation can be bound by an exchange no named person reviewed, recorded in a form no one can later produce.
Most contracting tooling was built for people signing documents, so it offers nothing that sits between an agent's acceptance and a binding obligation.
Why the record matters more than the negotiation
A negotiation that cannot be reconstructed is a negotiation you cannot defend. Subchain writes each proposal, counter, revision, approval and acceptance to an immutable log, so a settled agreement can be traced from first offer to final term, with the authorising person attached at each gate.
If a counterparty later disputes what was agreed, or a regulator asks on whose authority a commitment was made, the answer is retrieved rather than assembled from fragments. That legibility is the point: accountability should be visible in the record itself, not pieced together after a dispute has already begun.
What we can defend, and what is still unsettled.
What is not yet legally settled
We separate what we can defend from what we cannot. Automated contract formation is legally recognised in some jurisdictions, which removes the question of whether an agent-formed agreement can exist at all. What remains unsettled, in many of the places these agents run, is how far an agent's authority to bind extends and how liability is allocated when an agent errs.
Subchain does not resolve that for you. Where a term's enforceability or liability allocation is uncertain, the runtime is built to escalate the agreement for human legal determination rather than let it stand unexamined.
What this stage still has to prove
We are equally direct about demand. The need for a governed-negotiation layer is currently inferred from the integration and accountability pain around agent deployment, not yet confirmed by signed design partners. Turning that inference into evidence is the work of this stage.
If you already govern what agents commit the company to, the most useful thing you can tell us is where your approval authority sits today and what a dispute would require you to produce.
What a general counsel checks first.
Follow the governance layer as it forms.
Occasional updates for the people accountable for what agents commit the company to — how the gate, the record and the term library take shape. No frequent sending.