AMEG: governance built into the system, not written in a policy document
In short: AMEG is the governance layer of our research. It treats rules as part of the system itself: obligations, permissions and prohibitions are checked before every action an agent takes, and every change is written to a tamper-evident record. If an action cannot be shown to comply, it does not run, and it goes to a person.
The question it answers: How can autonomous execution remain bounded, auditable and accountable?
Governance as a document comes too late.
Governance is usually a document reviewed after the fact. When agents act in seconds, an after-the-fact review comes too late to prevent damage, and it is hard to reconstruct what happened.
How AMEG maps to the EU AI Act.
| EU AI Act | Usual approach | AMEG approach |
|---|---|---|
| Art. 9 Risk management | Periodic risk spreadsheets | Rules evaluated at runtime on every action |
| Art. 10 Data governance | Static data policies | Inputs constrained to valid, typed formats |
| Art. 11 Technical documentation | Manuals updated quarterly | Versioned architecture tied to every build |
| Art. 12 Record-keeping | Large unstructured logs | Append-only, tamper-evident record of every state change |
| Art. 14 Human oversight | Reviews after incidents | Automatic hand-over to a person when checks keep failing |
| Art. 15 Accuracy and robustness | Periodic scans | Agents hold no settlement keys; fixed clearing rails |
Built to travel: the same structure is designed to map onto other regimes, such as the NIST AI RMF and the UK framework, because they all ask for the same things: checked decisions, an audit trail and human escalation.
What the paper does not claim.
The containment result holds only under stated assumptions, and the tests used synthetic transactions. This is an architecture for meeting obligations, not legal advice or a certification.
Where it shows up.
"Limits we state openly" on How We Work, the EU AI Act section on Research, and the Government & public administration page.
Autonomous Micro-Enterprise Governance (AMEG)
How to cite: Rajagopal, R. K. (2026). Autonomous Micro-Enterprise Governance (AMEG). Secular Labs AME Architecture Series. https://doi.org/10.5281/zenodo.22718641
Frequently asked
Does using AMEG make a system EU AI Act compliant?
No single architecture does that on its own. AMEG gives you runtime controls and records that map to the Act's articles. Compliance still needs your legal review.
What is a tamper-evident record?
A log where each entry is linked to the one before it by a cryptographic hash, so any later change shows.