The Pulse
VisionX AI unveils AEGIS graph model and TraceX investigation platform
VisionX AI has unveiled AEGIS, a graph model designed to detect financial crime across accounts and wallets. Its TraceX blockchain investigation platform is being built on the model, with capabilities including reverse tracing planned.

VisionX AI has unveiled AEGIS, a graph model designed to detect financial crime by examining connections between accounts, wallets and transactions. The Singapore-based company also introduced TraceX, a blockchain investigation platform being built on the model, according to a press release issued on Nov. 18.
Source prnewswire.com
VisionX AI is a standalone company within Alpha Ladder Group. It combines its in-house graph model with large language models from established providers in what it calls a hybrid AI architecture.
Mapping financial networks
AEGIS is based on a Graph Transformer architecture and was pretrained on financial transaction graph data. The model's name stands for Attention-based Entity and Graph Intelligence System.
The model is designed to learn how wallets, accounts and entities connect, as well as how funds move between them. VisionX AI says it can score risk and recognise laundering patterns, using a network view to identify relationships that link activity across transactions.
In the hybrid architecture, AEGIS supplies network intelligence, while the language models handle natural-language interaction, reasoning and AI-assisted investigative workflows. The company describes the combination as a way to help investigators find suspicious fund flows and connections behind individual transactions.
In July 2026, the Financial Action Task Force warned that criminal networks exploit the borderless nature of virtual assets to commit fraud, evade sanctions and launder proceeds, according to the release. The task force found that illicit funds moved through channels including stablecoins, cross-chain bridges, decentralised exchanges and mixers, where they were mixed with legitimate flows.
TraceX remains in development
TraceX is being built for financial crime investigators and law enforcement agencies to analyse blockchain transactions and identify connections. The platform is designed to trace fund flows through linked addresses and transactions and to identify suspicious patterns across separate investigations.
Planned capabilities include reverse tracing, MegaMap network analysis, pattern recognition and case association. VisionX AI says the platform is intended to generate investigative leads using graph intelligence and AI-assisted analysis, supported by traceable evidence and human review.
Investigators will make decisions guided by TraceX recommendations, with each decision recorded. Chief Executive Officer Summer Yu said AI would indicate where investigators should look next, while their judgment and the evidence would determine the case.
From compliance platform to company
VisionX was originally incubated within MetaComp, Alpha Ladder Group's payments subsidiary, to address compliance and financial crime risks in regulated financial operations. It has since developed into a broader financial crime and compliance platform spanning traditional finance and digital assets.
The platform supports the compliance intelligence infrastructure of MetaComp's StableX Agentic Network, which launched as the StableX Network in November 2025. It also includes VisionX KYT, which consolidates blockchain risk intelligence from independent analytics vendors.
VisionX KYT draws on those vendors' data for 1 billion wallets across 25 chains, the company says. Another product, Web2.5 TxM, brings together identity, behavioural and network intelligence for transaction monitoring across traditional finance and digital assets.
Source
Related stories
- Microsoft publishes quantized Aion model for WebGPU
- Gemini Notebook can turn research sources into interactive sites
- Google rolls out Gemini 4 Neon to developers and paid subscribers
- Google opens global waitlist for experimental Stitch Loop agent
- Google details Sashiko's agentic reviews of Linux patches