No Blockchain, No AI?
What? Surely not.
Yet, the chair of America's top derivatives regulator said something on a podcast last week that I didn't expect.
"You can't have AI without blockchain."
CFTC Chair Michael Selig told The Pomp Podcast that blockchain timestamps and on-chain identifiers may be the verification layer that distinguishes real media from AI-generated content (CoinTelegraph, April 2026). In a world awash with synthetic content, blockchain could be the authenticity infrastructure.
He also pushed back against over-regulation: "I'm concerned we over-regulate and strangle some of the technology." On AI agents specifically, his view was that enforcement should target people engaging in financial activity, not developers building the tools.
That's a seismic mental shift.
For years, the regulatory dialog in Washington focused on whether crypto was a security, a commodity, or a liability. A key regulator is now saying that blockchain is essential plumbing for the next generation of the internet.
The institutions building AI trading systems, AI compliance tools, and AI-generated research products must eventually solve for content authenticity and auditability. Blockchain is being positioned as one possible answer.
Every thought from a regulator doesn't necessarily become reality. Still, when the head of US derivatives oversight considers blockchain as AI infrastructure, then the investment thesis has certainly gone well beyond crypto token prices.
Interesting times!
What Selig Actually Said, and Why It's Different from the Usual Crypto Pitch
Let me be specific about what Selig said on The Pomp Podcast, because the framing matters enormously.
He wasn't arguing that blockchain is a great store of value. He wasn't talking about tokenization of financial assets. He was talking about provenance: the ability to establish a reliable, tamper-resistant record of when something existed and who created it.
His specific argument: as generative AI becomes capable of producing photorealistic images, convincing video, and written content indistinguishable from human output, the existing tools for verifying authenticity, metadata, digital watermarks, AI detection models, are in a losing arms race with the generation capabilities. He positioned blockchain as a potential foundation for a verification layer that's harder to fake retroactively.
The core logic: an immutable, timestamped on-chain record of a piece of content's existence at a specific moment, by a specific creator, on a specific chain. You can't alter the past on a public blockchain. That's a different kind of guarantee than metadata, which can be stripped and replaced.
This is not a crypto pitch. It's a statement about internet infrastructure.
The Content Authenticity Problem Is Already Costing Real Money
Here's why this isn't academic.
In early 2024, a finance worker at a Hong Kong company authorized a $25 million wire transfer after a video call with what appeared to be his company's CFO and several colleagues (CNN, February 2024). It was a deepfake. Every person on the call was AI-generated. The $25 million was gone.
That was 2024, with generation capabilities that were already impressive but not at their current level. The attack surface has expanded significantly since.
Financial executives, particularly at institutions where a single executive's verbal authorization carries operational weight, are increasingly high-value targets for AI-generated impersonation. A convincing deepfake of a CEO instructing a wire transfer. A synthetic audio clip of a CFO approving a transaction. An AI-generated written instruction that passes style-matching checks.
Existing verification tools are struggling. Metadata can be stripped. Watermarks can be removed or mimicked. AI detection models are in an arms race with generation models: every time a detector improves, generation models adapt. The cat-and-mouse dynamic doesn't obviously favor the defenders.
Blockchain-based provenance doesn't solve the detection problem. But it creates a positive verification alternative: instead of trying to detect whether something is fake, you verify whether it has a legitimate provenance record. The absence of a valid on-chain record doesn't prove something is fake, but its presence proves it was registered at a specific time by a specific entity.
That's a meaningful capability, even if it's not a complete solution.
What Blockchain-Based Content Provenance Actually Looks Like
This isn't a theoretical proposal. Work is already underway.
The Content Authenticity Initiative (CAI), backed by Adobe, Microsoft, Sony, and others, is developing standards under the C2PA framework (Coalition for Content Provenance and Authenticity) (Content Authenticity Initiative, 2024). The idea is to embed cryptographic signatures at the point of content creation, establishing a chain of custody from creator to consumer.
Blockchain is one component of that stack. At content creation, a cryptographic hash of the content gets registered on-chain with a timestamp. The verification layer can then confirm: this specific content existed in this specific form on this specific date, registered by this specific entity. Any modification to the content changes the hash, breaking the provenance chain.
For financial institutions specifically, this has direct applications. AI-generated research reports, AI-produced compliance summaries, AI-drafted client communications: any output where authenticity and auditability matter could carry a verifiable provenance record. The audit trail isn't just who wrote it, but what tool produced it and when.
Regulators who care about record-keeping obligations, which is most financial regulators, should find this interesting.
The Honest Limits of What This Solves
I want to be intellectually honest here, because the enthusiasm around blockchain-as-provenance-layer can outrun what the technology actually delivers.
Blockchain timestamps can tell you when something was registered. They cannot tell you whether what was registered is true.
An AI-generated image registered on-chain is still AI-generated. A deepfake with a provenance record is still a deepfake. The on-chain record proves the registration exists, but it doesn't validate the content's authenticity. If an attacker registers synthetic content before using it in a fraud, the provenance record becomes evidence of the fraud's premeditation, not a barrier to it.
The more precise claim is this: blockchain provenance makes retroactive fabrication harder. If the legitimate CEO's communications all carry on-chain provenance records, and a synthetic communication doesn't, or carries a provenance record created after the fact, the absence or timing mismatch becomes a detectable signal. That's useful. It's not a complete defense.
The gap between "useful verification layer" and "complete solution" is significant, and anyone building systems around blockchain provenance needs to be clear-eyed about it. Selig didn't claim it was a complete solution. He positioned it as a necessary component. That's probably the right framing.
The Regulatory Philosophy Shift: From Defensive to Essential
For years, the crypto industry's posture in regulatory conversations was essentially defensive. Tokens aren't securities. DeFi isn't a broker-dealer. NFTs aren't financial instruments. You're always reacting, always carving out exceptions, always at the mercy of how regulators choose to categorize you.
Selig's framing inverts the argument. He's not treating blockchain as a financial product that needs classification. He's treating it as infrastructure society needs to function in an AI-dominated world. That's a fundamentally stronger political and regulatory position.
We've seen this script before with the internet. TCP/IP wasn't regulated as a financial product. The web wasn't subject to securities law. They were treated as infrastructure, which attracted different capital and different institutional adoption patterns. If blockchain gets even partially recategorized as critical internet infrastructure rather than as a financial innovation, the adoption implications are significant.
On AI Agents and Enforcement Targeting People, Not Developers
Selig's other comment on the podcast deserves its own moment of attention.
His view on AI agent enforcement: go after the people engaging in financial activity, not the developers building the tools. Liability should attach to the human directing the agent, not to the programmer who wrote the agent's code.
This is significant for how institutions think about AI deployment. The implicit question whenever you deploy an AI system in a regulated context is: if this goes wrong, who's liable? If the answer is "the developer," you get a chilling effect on development. Firms won't build tools if building them creates open-ended liability exposure. If the answer is "the person deploying and directing the tool," you get a framework where the institution carries the responsibility for how it uses the technology.
That's actually more consistent with how we think about professional liability in other domains. A law firm is liable for what its lawyers do, not for what legal research tools produce. A bank is liable for what its traders do, not for what trading software outputs.
Applying that logic to AI agents, especially in financial contexts, creates a workable framework where institutions can deploy AI systems responsibly, with clear accountability structures. It's not a blank check; it's a framework that assigns liability to the right place.
What This Means for Institutional Investment in Blockchain
The investment thesis for blockchain technology has gone through several iterations.
First it was "Bitcoin is digital gold." Then "Ethereum enables smart contracts." Then "DeFi will replace TradFi." BlackRock's Fink put it as being at the 1996 moment of the internet, which is accurate and unsettling in equal measure. Each framing attracted different capital with different time horizons and different risk appetites.
The framing Selig is gesturing toward is different from all of them: "blockchain is infrastructure the internet needs to function properly in an AI-dominated world." That attracts infrastructure investors, not speculative investors. It attracts institutional capital that's looking for durable utility, not token price appreciation.
Infrastructure investors have longer time horizons, lower return expectations, and more patient capital. They also tend to be the investors who show up when a technology transitions from experimental to essential. The internet's infrastructure phase, fiber, data centers, routing equipment, attracted a different class of capital than the browser wars did.
If blockchain's provenance use case gains genuine traction, and the combination of deepfake proliferation and regulatory attention makes that more likely than it was two years ago, the capital that follows will look different from what's been in the space historically. That's probably healthy.
My Genuine Reaction
When I first encountered Selig's comments, my instinct was mild skepticism. Regulators don't typically become cheerleaders for the technologies they oversee. "Blockchain is essential infrastructure" from the head of US derivatives oversight felt almost too good to be true.
But the reasoning is sound, even if the outcome is uncertain. The content authenticity problem is real and growing. The existing tools are genuinely inadequate. Blockchain's immutability and timestamping are useful for the provenance problem, even if they don't solve it completely.
The gap between Selig's observation and a functioning blockchain-based content provenance ecosystem is still significant: technically, institutionally, and politically. Regulators saying something on a podcast and building frameworks around it are different things.
But direction matters. When the head of US derivatives oversight frames blockchain as AI infrastructure rather than as a financial product to be regulated, the conversation has moved somewhere new. That's worth tracking. AI agents are already transacting autonomously in crypto, and the compliance frameworks for that reality don't exist yet.