Elliptic

Learning Objectives

• Understand how blockchain-analytics platforms screen on-chain transactions for financial-crime risk, and where genuine machine learning fits in that stack.
• Identify Elliptic's core products — Lens, Investigator, Discovery, and the Elliptic Copilot AI assistant — and what each one does.
• Evaluate the honesty questions around crypto-compliance AI: probabilistic risk scoring, undisclosed accuracy metrics, and the financial-privacy trade-offs of deanonymizing blockchain activity.

What Is Elliptic?

Elliptic is a blockchain-analytics and crypto-compliance company that screens on-chain transactions for financial-crime risk — anti-money-laundering (AML), sanctions exposure, and fraud — and provides investigation and intelligence tooling to crypto exchanges, banks, and governments. The core problem it addresses is that public blockchains are pseudonymous: anyone can see the transactions, but nobody can natively see who is behind an address or whether the funds passed through a sanctioned entity, a scam, or a darknet market. Elliptic builds the labeling, tracing, and risk-scoring layer that lets a regulated institution answer "where did this money come from, and can I safely touch it?" before onboarding a customer or clearing a transfer.

Founded in 2013 and headquartered in London, United Kingdom (with offices in New York, Singapore, and Tokyo), Elliptic was started by Dr. Tom Robinson, Dr. James Smith, and Dr. Adam Joyce. The company is privately held and has raised roughly 224 million dollars in total. Its Series D of 120 million dollars closed around May 12, 2026 at a 670 million dollar valuation, led by One Peak with participation from Nasdaq Ventures, Deutsche Bank, and the British Business Bank. In September 2025 Elliptic became the first blockchain-analytics firm backed by four globally systemic banks — HSBC, JPMorgan Chase, Santander, and Wells Fargo — an unusually deep institutional endorsement for the sector.

Core Capabilities

Lens — Real-Time Transaction and Wallet Screening

Lens is Elliptic's real-time, multi-asset screening product. It checks wallets and transactions across more than 60 blockchains against risk signals — sanctions lists, known illicit addresses, and exposure to high-risk services — so an exchange or bank can score a counterparty before or during a transfer. This is the workhorse compliance product most customers touch daily.

Investigator — Cross-Chain Fund Tracing

Investigator provides single-click cross-chain fund tracing. When funds move across bridges and between blockchains, an analyst can follow the trail through Elliptic's transaction graph rather than manually reconstructing it. This is the "holistic screening" capability in practice, and it is primarily graph and data engineering rather than a learned model.

Discovery — Crypto-Business and VASP Due Diligence

Discovery screens crypto businesses and Virtual Asset Service Providers (VASPs) for counterparty due diligence during onboarding. Instead of scoring a single transaction, it assesses whether an entire exchange or service is a safe counterparty, which banks use when deciding whether to bank a crypto firm.

Elliptic Copilot — The AI Assistant

Elliptic Copilot is the flagship AI feature: an LLM assistant embedded across the screening workflows. It generates screening summaries, entity descriptions, risk-graph narratives, draft Suspicious Activity Reports (SARs), and enhanced-due-diligence assessments — turning raw graph output into readable analyst-facing text. It anchors Elliptic's stated "agentic" product roadmap, where more of the investigation workflow is drafted by the model and reviewed by a human.

Elliptic Data Set and MIT Research

Elliptic has a genuine machine-learning research pedigree. It partnered with MIT and IBM to apply deep learning to a public dataset of more than 200,000 Bitcoin transactions to detect illicit activity. That widely cited "Elliptic Data Set" has become a standard benchmark in academic anti-money-laundering research, and it grounds the company's claim to real ML work rather than pure marketing.

Strengths

• Deep institutional backing: Backed by HSBC, JPMorgan Chase, Santander, Wells Fargo, Deutsche Bank, and Nasdaq — an unusually strong bank-side endorsement for a chain-analytics vendor.
• Cross-chain coverage: Holistic screening follows funds across and between more than 60 blockchains, a genuine differentiator as illicit flows increasingly hop chains.
• Genuine ML research pedigree: The MIT and IBM collaboration and the public Elliptic Data Set show real machine-learning work, not just a rules engine with an AI label.
• Copilot productivity layer: The LLM assistant meaningfully speeds up analyst reporting by drafting summaries, narratives, and SARs.
• Scale: Company-reported figures cite more than 700 customers across roughly 30 countries and screening of more than 1 billion transactions per week.

Limitations and Considerations

• AI-versus-heuristics is blurred: Much of the core — holistic screening, cross-chain tracing, risk scoring — is graph traversal, rules, and address-clustering heuristics, not learned models. The genuinely ML parts are the illicit-activity classifiers and Copilot.
• No published accuracy metrics: Elliptic discloses efficiency gains but no precision, recall, or false-positive rates and no model-validation methodology, so buyers cannot independently judge screening quality.
• Probabilistic scoring flags innocents: Blockchain risk scoring is probabilistic — legitimate users can be flagged simply by proximity to a flagged address, a guilt-by-association failure mode.
• Automation-bias and hallucination risk: Because Copilot drafts SARs and entity summaries, analysts who rubber-stamp AI output risk filing inaccurate or hallucinated conclusions; a human must verify.
• Financial-privacy trade-offs: Like all chain-analytics firms, the business deanonymizes blockchain activity for banks and governments, raising real financial-privacy concerns.
• Smaller than the leader: Elliptic is one of the top three Western vendors alongside market leader Chainalysis (larger, at about an 8.6 billion dollar valuation) and TRM Labs, but is smaller than Chainalysis by scale.

Best Use Cases

Getting Started

1. Identify your compliance need — real-time screening (Lens), investigation and tracing (Investigator), or counterparty due diligence (Discovery) — since Elliptic is sold as an enterprise suite.
2. Contact Elliptic through elliptic.co for a demo and enterprise pricing; access is quoted on request rather than self-serve.
3. Integrate screening into your onboarding and transaction-monitoring workflows via the API, and decide where Copilot drafts feed your analysts.
4. Establish a human-review policy for any Copilot-generated summary or Suspicious Activity Report so probabilistic scores and AI drafts are verified before action.

Key Takeaways

• Elliptic is a London-based blockchain-analytics and crypto-compliance company, founded in 2013, that screens on-chain transactions for financial-crime risk across more than 60 blockchains.
• Its flagship AI feature is Elliptic Copilot, an LLM assistant that drafts screening summaries, risk narratives, and Suspicious Activity Reports; its cross-chain "holistic screening" is mostly graph analysis rather than machine learning.
• The company has a real ML research pedigree through its MIT and IBM collaboration and the public Elliptic Data Set of more than 200,000 Bitcoin transactions.
• Honesty caveats matter: no published accuracy metrics, probabilistic scoring that can flag innocent users, automation-bias risk when analysts rubber-stamp Copilot output, and financial-privacy trade-offs from deanonymizing blockchain activity.
• The genuine AI in crypto is this intelligence-and-compliance layer — not hype-driven "AI token" coins; Elliptic sits among the top three Western vendors alongside the larger Chainalysis and TRM Labs.