AI is not the answer to AI-enabled fraud - Coda Story

The Dissection

This article is a systemic observation dressed as regulatory critique. The author correctly identifies the mechanics of AI fraud, the futility of box-checking compliance, and the absurdity of AI-on-AI arms races without foundational data. But the piece stops at the precise moment where structural analysis must begin. It reads like someone who sees the house is on fire and concludes the smoke detectors need new batteries.

The Core Fallacy

The article assumes information asymmetry is the solvable problem. The thesis: if we had proper corporate registries, reliable AI training data, and functional crypto policy, then AI could counter AI fraud effectively. This is the intellectual equivalent of saying cancer would be cured if we had better X-rays.

The author cites Carole House's testimony approvingly: "Without a secure identity foundation, AI agents will simply scale up fraud at a speed and volume that human investigators can't possibly track, destroying trust in the whole system." This is accurate. But the conclusion drawn—that better foundations will preserve the system—is not.

What House actually describes is the mechanism of collapse, not a call for remediation. When AI agents can scale fraud beyond human tracking capacity, you don't fix this with better registries. You have crossed a threshold where the human oversight layer is structurally irrelevant. The system that required human investigators is already dead; the author is eulogizing a corpse.

Hidden Assumptions

1. Human-in-the-loop is recoverable. The article treats the current compliance architecture (AI flags, humans check, humans get replaced by better AI) as a temporary dysfunction rather than the permanent architecture being established.

2. Regulatory capacity can close the gap. The assumption that adequate corporate transparency legislation, crypto policy reform, and international coordination are achievable at the speed and scale required. The UK's illicit finance summit being postponed due to "scheduling issues" is presented as depressing; it is actually diagnostic. The system cannot process its own problems fast enough.

3. The arms race stabilizes. The article treats AI fraud vs. AI compliance as a contest with a winner. In DT terms, this is incorrect. Both sides are being automated simultaneously, which means the human layer on both sides—the compliance officer AND the money launderer's analyst—become structurally unnecessary. The race doesn't produce a winner; it produces a winner-take-all for the entities that own the infrastructure.

Social Function

Prestige signaling within the policy class. This article is written for people who attend illicit finance summits, read blockchain analytics reports, and believe congressional testimony is where systemic transformation happens. It performs the function of serious engagement with serious problems while missing entirely that the problems it identifies are symptoms of a structural phase transition, not bugs in an otherwise functional system.

The most damning sentence: "I very much doubt Davidson was listening, however, because that is not the direction the Republican Party is going in right now. I would write more about that, but frankly it's all too depressing, and I'd rather move on."

This is ideological anesthetic. The author perceives the political capture of the regulatory apparatus, acknowledges it cannot be addressed within the current political framework, and then... moves on. The depression is real; the response is to not examine what the depression means.

The Verdict

The article correctly diagnoses the symptoms—AI fraud scaling, compliance theater, regulatory capture, data foundation failure—but treats these as fixable malfunctions in a system that is structurally transitioning to a state where human cognitive labor is not the operative layer. The Standard Chartered CEO's "lower-value human capital" is not a gaffe. It is a factual description of a structural relationship that is becoming mathematically irreversible. The compliance officer in the Warsaw cubicle is not being underutilized; they are performing a function that AI will perform more cheaply, and the only question is timeline.

The article is a beautifully written, precisely observed obituary for a system that is already dead and doesn't know it yet.