CUBE3: Real-Time Fraud Prevention Across Web2 and Web3

CUBE3 — now operating under the brand CUBE AI — is a Palo Alto-based security company that applies machine learning to one of Web3's most persistent problems: fraudulent and malicious transactions executing on-chain before anyone can stop them. Founded in 2021 and built by a team that spans serial entrepreneurship, AI research, and executive-level financial crime prevention, the platform aims to catch threats before the funds move rather than after.

The Core Problem CUBE3 Addresses

On-chain fraud has historically been a post-mortem exercise. Protocols are exploited, funds are drained, and analysts trace the damage afterward. Smart contract audits catch code vulnerabilities before deployment but cannot evaluate the intent or history of the specific addresses that interact with those contracts at runtime. CUBE3 targets this gap: the moment a transaction is submitted and before it is finalized.

The company's broader thesis is that most damaging crypto scams do not originate entirely on-chain. Pig butchering schemes, investment fraud, ransomware, and sextortion all begin in Web2 environments — social media, messaging apps, phishing sites — before the victim ultimately sends cryptocurrency. Tracking only the blockchain layer captures the symptom but not the cause.

How the Platform Works

CUBE3's risk engine ingests two parallel data streams. The first is real-time blockchain transaction data across supported networks. The second is Web2 signals: social media platforms, dark web forums, known phishing infrastructure, compromised address databases, and online scam activity. Proprietary machine learning models fuse these streams and generate a dynamic Risk Score from 1 to 100 for every transaction.

The score is broken down across three threat categories:

• Cybersecurity risk — known exploit patterns, malicious contract code, attack vectors
• Fraud risk — scam address associations, behavioral anomalies, velocity indicators
• Compliance risk — government sanctions lists, money laundering patterns, regulatory watchlists

Factors evaluated per transaction include origin and destination address history, transaction velocity, amount patterns, and behavioral correlation against CUBE3's global threat database.

CUBE3 Protect: Smart Contract RASP

The most technically differentiated product is CUBE3 Protect, which implements Runtime Application Self-Protection (RASP) directly at the smart contract level. Developers integrate it by inheriting from CUBE3's base contracts and applying protective modifiers to the specific functions they want to secure. When a user submits a transaction to a protected contract, CUBE3's risk engine evaluates it in real time. If the risk score exceeds the configured threshold, the transaction is automatically reverted before it executes — without pausing the protocol or requiring manual intervention.

This architecture means protection is embedded in the contract itself rather than sitting as an external monitoring layer that can only alert after the fact. CUBE3 Protect currently operates across Ethereum, Polygon, Arbitrum, BNB Chain, and Avalanche.

CUBE3 Inspector and Solana Support

Alongside Protect, CUBE3 offers Inspector — a risk assessment tool for evaluating individual wallet addresses. Inspector takes an address and returns a risk profile built from CUBE3's combined Web2/Web3 threat intelligence.

Inspector's supported networks have been expanded beyond EVM chains to include Bitcoin, Litecoin, Tron, and Solana. This makes CUBE3 one of a limited set of Web3 security providers offering pre-transaction address-level risk screening for Solana wallets and counterparties. While the full RASP smart-contract integration is currently an EVM-only feature, Solana-native protocols and payment platforms can use Inspector via API to screen incoming addresses and flag high-risk counterparties before processing transactions.

Phantom Ledger: Closed Intelligence Briefs

The third product, Phantom Ledger, targets compliance and fraud teams at financial institutions rather than on-chain developers. It is a private, curated intelligence brief covering active fraud infrastructure — scam packaging, laundering routes, and monetization flows — shared through periodic closed-door briefings and detailed written breakdowns. Access is application-gated; CUBE3 individually reviews each submission to verify recipients are appropriate professionals in fraud prevention, risk management, or trust and safety roles.

Team

The company's founding team brings an unusual combination of consumer cybersecurity scale and institutional financial crime expertise.

Einaras von Gravrock (CEO) is a serial founder whose previous company, CUJO AI, built AI-powered network security for broadband operators and was ultimately adopted to protect more than 60 million homes following acquisitions by Comcast and Charter. He subsequently co-founded CAST.AI, a cloud optimization platform serving Fortune 500 clients, and has been recognized by the World Economic Forum and Goldman Sachs.

Jonathan Anastasia (President) spent more than a decade at Mastercard, reaching EVP level where he led Global Crypto Services and Financial Crimes. He managed a team of over 400 across 15 markets, shaping fraud strategy and regulatory compliance for one of the world's largest payment networks.

Attila Marosi-Bauer serves as Co-Founder and Chief Research Officer. The company employs 65 or more engineers and AI specialists across its security and research functions.

Funding and Investors

CUBE3 has raised approximately $27.9 million across four rounds from seven institutional investors:

• An early pre-seed round of roughly $3.8 million, backed by Dispersion Capital and four co-investors
• A seed round of approximately $8.4 million in August 2023, led by Blockchange Ventures
• A Series A tranche of roughly $6.1 million in June 2024
• A Series A extension of approximately $9.2 million in December 2025

Other disclosed investors include Symbolic Capital, Hypersphere Ventures, ICLUB, and TA Ventures.

Ecosystem Position and Recognition

CUBE3 was selected for Mastercard's Start Path Blockchain and Digital Assets Program, making it the first blockchain cybersecurity company admitted to the program since it was established in 2014. The company also went through the Berkeley Blockchain Xcelerator. It holds ISO 27001 certification and maintains GDPR, CCPA, and SOC 2 compliance — a credential set that matters to the regulated financial institutions CUBE3 targets as customers.

The platform reports protecting more than 100 organizations. Notable threat detections include identifying and preventing an attempted $28 million exploit against Penpie, a $2.28 million phishing attack targeting Aave users, and flagging over $17 million in rug pull activity within a 48-hour window. In April 2025, CUBE3 announced a partnership with Predicate to deepen fraud prevention coverage for blockchain transaction flows.

Competitive context: CUBE3 sits in a market alongside better-capitalized incumbents including CertiK ($296 million funded), Elliptic ($224 million), and Solidus Labs ($83 million). Its differentiator is the real-time, pre-execution RASP integration at the smart contract level combined with cross-layer Web2 signal ingestion — an approach that targets the full fraud lifecycle rather than either code audits or post-transaction analytics alone.