The Elliptic Standard

As tokenized deposits, stablecoin settlement and other on-chain activity reach mainstream banking, the volume and speed of that activity is starting to outrun case-by-case human review. Institutions are turning to agentic AI to keep up. Accountability for every resulting decision stays exactly where it was.

The Elliptic Standard sets out eight principles for agentic on-chain risk, written for the compliance, risk and policy leaders who will be accountable for these systems and the teams who will run them.

Why it matters

Stablecoin transaction volumes reached $33 trillion in 2025, up 72% year on year. Autonomous agents are already originating, authorizing and settling a growing share of transactions, and illicit actors have the same access to AI without the constraints regulated institutions face. Detection built around after-the-fact review cannot keep pace.

What it covers

The eight principles run from the quality of the data an agentic system reasons over, through the transparency and validation of the models behind it and the safety risks that come with agentic AI itself, to the institution's own control over thresholds, escalation and intervention. They also cover operational resilience when the agentic process is unavailable, and what the risk role becomes when process execution automates.

Each principle sets out what a system has to do, and what an institution should require of it, and they map onto frameworks institutions already run, from third-party due diligence to model risk governance.

Elliptic published the Standard in the open, written so it can be applied to any provider, and has adopted the principles in its own development.

Download to read the full text.