Elliptic launches Decode AI agent for natural-language on-chain risk queries

Elliptic launches Decode AI agent for natural-language on-chain risk queries

N
News Editor
2026-09-18 03:07:07
Blockchain risk analytics firm Elliptic has introduced Decode, an AI agent built on its own platform to answer questions about crypto wallet addresses and risk exposure in seconds using natural language. The company says the tool is designed to speed up investigations that previously required analysts to rely on specialist staff to write database queries, a process that could slow casework. Decode can interpret prompts, show the query it generates, and return ranked tables alongside short summaries. It also supports batch analysis across multiple addresses, helping identify counterparties, repeated transactions, and common destinations. Elliptic said the product is built on its Elliptic Standard framework, with an emphasis on transparency and verifiability, and that each response includes the reasoning behind the result. The company added that human oversight remains in place, with AI limited to suggesting follow-up leads while final investigative decisions stay with analysts. Decode is now available to government customers, and Elliptic plans to open it to private companies and financial institutions later this year.

Blockchain risk data analytics firm Elliptic has launched Decode, an AI agent built on the company’s platform that is designed to answer questions about crypto wallet addresses and their risk exposure in seconds through natural-language prompts, while providing traceable data support.

Natural-language prompts replace manual query work

Elliptic said financial analysts have traditionally needed support from data specialists to write queries, a setup that can slow investigations. Decode is meant to remove part of that bottleneck by letting analysts ask questions directly in plain language.

Examples cited in the report include prompts such as asking who the largest counterparty to an exchange was by trading volume over the past six months, or asking who controls a given address and what risks it is exposed to. After receiving a prompt, the system analyzes the request, displays the query it generates, and then produces ranked tables and a short summary.

For large batches of addresses, Decode can also identify counterparties and repeated transactions automatically, reducing work that would otherwise require analysts to piece together results manually.

Built on the Elliptic Standard framework

Jackson Hull, Elliptic’s chief technology officer, said the market is crowded with AI agents that claim to be trustworthy but still fall short of what analysts need for decision-making. He said Decode was developed under the company’s Elliptic Standard framework to provide model transparency and verifiability.

According to Elliptic, every answer is delivered in natural language and includes the logic behind the result, so conclusions can be tied back to a source and supporting evidence rather than appearing as unsupported output.

Batch analysis with human oversight

Decode can run queries across a group of addresses or a broader case set to surface counterparties, common destinations, and repeated patterns. Elliptic said this allows analysts to avoid the slower, more labor-intensive process used in traditional workflows.

The company also said the tool includes a human oversight mechanism. AI is limited to suggesting follow-up directions, while final investigative decisions and case closure remain with human analysts. Elliptic added that Decode draws on a labeled address database accumulated since 2013 to improve data accuracy.

Government rollout first, private-sector access later

Decode is currently available to government agency customers, according to the report. Elliptic plans to make the product available to private enterprises and financial institutions later this year, with the stated aim of helping audit teams in their decision-making process.

ABMedia’s summary of the launch also said the product combines tens of billions of labeled addresses and attribute data accumulated since 2013. It described the technical design as one that preserves verifiable query steps and addresses a common weakness in general-purpose AI systems: the lack of clear data provenance.

This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.
2900

Disclaimer:

The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.

Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.