The rapid transition toward autonomous, machine-led transactions is fundamentally altering the landscape of financial crime and on-chain risk management. As autonomous agents begin to originate, authorize, and settle transactions at speeds exceeding human review capabilities, traditional backward-looking forensic models are becoming increasingly inadequate. Elliptic, a global on-chain risk provider, has responded to this inflection point by publishing "The Elliptic Standard," a framework consisting of eight core principles designed to govern agentic on-chain risk. This move seeks to address a critical gap in financial infrastructure where the speed of illicit AI-driven activity threatens to outpace existing compliance architectures, potentially undermining the stability of mainstream on-chain finance.
The Surge in Agentic Transactions and AI-Linked Losses
The shift toward an "agentic" economy is driven by three distinct market forces that are rapidly scaling the volume and complexity of on-chain activity. First, the rise of autonomous agents as transaction counterparties is projected to push agent-facilitated consumer spending to $3.35 trillion by 2030, according to research from PHD and WARC. Second, the scale of the threat is accelerating; the FBI's Internet Crime Complaint Center recorded $893.3 million in adjusted 2025 losses tied to an identified AI nexus. Third, the mainstreaming of on-chain finance is evidenced by stablecoin transaction volumes, which reached $33 trillion in 2025.
Elliptic argues that current blockchain analytics are built for a pre-agentic world, relying on investigation-led models that occur after a transaction has settled. This "forensics-first" approach is increasingly viewed as insufficient for a system where risk decisions must be made before settlement to prevent systemic issues. The company is positioning its new standard as a requirement for moving from simple automation to compliant, explainable agentic intelligence. By emphasizing real-time detection and defensibility, the framework aims to provide a roadmap for institutions to manage the risks inherent in a landscape where machines, rather than humans, are the primary drivers of transaction volume and velocity.
Establishing a Defensible Framework for Agentic Compliance
To address the limitations of manual review, The Elliptic Standard outlines eight specific principles intended to create a regulator-credible risk intelligence system. These principles include ensuring exceptional data quality that is real-time and comprehensively attributed, and maintaining model transparency to avoid "black box" scenarios. The standard explicitly addresses AI safety concerns, such as drift, hallucination, bias, prompt injection, and data poisoning. Furthermore, it advocates for foundation model agnosticism to mitigate concentration risk and ensure flexibility across different jurisdictional obligations.
Crucially, the framework emphasizes that human oversight must be "by design" rather than an afterthought. This involves building the ability to define, audit, and override agent decisions directly into the system architecture. The standard also calls for business resilience through defined fallbacks when agents are unavailable and requires deep employee training to ensure staff can interrogate and challenge agentic decisions. Elliptic, which developed the standard with input from regulated institutions, has committed to applying these eight principles to its own platform, signaling a push for industry-wide adoption of these accountability measures to meet the growing expectations of global regulators.
Key Takeaways
- Agent-facilitated consumer spending is projected to reach $3.35 trillion by 2030, according to PHD and WARC research.
- The FBI's Internet Crime Complaint Center reported $893.3 million in adjusted 2025 losses linked to an identified AI nexus.
- Stablecoin transaction volumes reached $33 trillion in 2025, highlighting the mainstreaming of on-chain finance.
FinanceInsyte's Take
In our view, the publication of The Elliptic Standard marks a critical acknowledgment that the "detect and investigate" era of blockchain compliance is effectively over. As stablecoin volumes hit $33 trillion and autonomous agents begin to dominate transaction flows, the window for human-in-the-loop intervention is closing. This shift necessitates a move toward "preventative" rather than "forensic" compliance.
The strategic importance of the "explainability" requirement cannot be overstated; for institutional players, an agent that acts without a traceable rationale is a regulatory liability, not an asset. By advocating for model agnosticism and human oversight by design, Elliptic is attempting to set the baseline for what "responsible AI" looks like in the capital markets. For C-suite executives, the takeaway is clear: the transition to agentic finance requires a complete overhaul of risk architecture, moving away from legacy oversight toward real-time, defensible, and highly configurable automated systems.
Questions & Answers
How does the rise of autonomous agents impact the scale of financial crime risk?
Autonomous agents can originate and settle transactions at a volume and speed that human teams cannot manually review. This creates a vulnerability where illicit actors using AI can exploit the velocity of on-chain finance to evade traditional, backward-looking detection models.
What are the primary regulatory concerns regarding the use of AI in transaction monitoring?
Regulators are expected to focus on whether an institution can explain why an agentic system made a specific decision. The Elliptic Standard addresses this by emphasizing model transparency, validation, and the requirement that human oversight be built into the system architecture from the start.
Why is "foundation model agnosticism" a critical principle for financial institutions?
Being locked into a single AI model creates concentration risk. Agnosticism allows firms to remain flexible, ensuring they can meet varying jurisdictional obligations and adapt to new technological developments without being tethered to a single provider's limitations.
What is the difference between "automation" and "compliant agentic intelligence" according to Elliptic?
According to Elliptic CEO Simone Maini, automation is simply an agent acting fast, whereas compliant agentic intelligence requires the ability to explain the reasoning behind an action. A system must be transparent, explainable, and subject to human override to be considered compliant.
Source: Businesswire
