Something is being rehearsed in Washington that most crypto traders are not watching, and it will hit our market before it hits the AI labs. The U.S. government is refusing to write new laws for artificial intelligence. Instead, it leans on decades-old liability doctrine — state tort law, the FTC Act, Section 230 — and lets the courts do the legislating.
That is the exact playbook it has run against us for five years. The SEC did not write a token law. It resurrected a 1946 Supreme Court test and pointed it at our order books. If you want to know where crypto regulation goes next, watch what happens to AI. The AI liability fight is the preview. Crypto is the sequel, and the sequel is already cast.
Let me be precise, because the phrase "old liability laws" hides an entire architecture. Under U.S. federalism, there is no single AI statute. There is a patchwork: 50 state tort regimes, Section 5 of the FTC Act covering unfair or deceptive conduct, Section 230 shielding platforms, and sectoral rules like the Fair Credit Reporting Act and Title VII that catch automated decisions sideways. The report itself is thin — three facts, no statute names, no defendants. But the thinness is the story. A regulator that had a plan would have published one.
The regulatory philosophy is "innovation first, liability later." No pre-approval. No licensing gate. Risk gets priced after the harm, in a courtroom, through damages. That is light-touch regulation in name and litigation-driven regulation in practice. Brussels chose the opposite — the EU AI Act grades risk before deployment and demands compliance upfront. Beijing chose a third way: mandatory registration of generative models before they ship. Three philosophies, three compliance stacks, and a multinational model builder must satisfy all of them at once.
The uncomfortable truth buried in the reporting: experts cannot agree whether AI liability should be technology-neutral or technology-specific. Translation — do we judge a model the way we judge a negligent driver, or do we admit a system nobody can explain needs its own rules? That disagreement is not academic. It is the same fork crypto has been stuck on since 2017.
Here is where my background matters. I have spent eighteen years watching regulators decide whether to apply old categories to new instruments, and I have paid tuition for the lesson. In 2017, at 25, I put $50,000 into four unaudited ICOs because the APYs were beautiful and the audits were not. Three rugged. I lost 80% of that book in a single correction. I traded hope for logic when the NFT bubble burst two cycles later, and the logic I settled on is simple: the legal category a technology gets assigned determines its market structure, its liquidity, and who survives.
Apply that lens to AI and the stakes sharpen. The central legal question — whether a software model counts as a "product" under product liability law — is the identical question that decides whether a token is a security. Both hinge on whether courts treat code as a good or as expression. If a model is a product, strict liability attaches, no fault required, and the deep-pocketed developer eats the loss. If algorithm output is speech, the First Amendment shields it and plaintiffs must clear a far higher bar.
Now watch the money follow the doctrine. When liability is uncapped and unpredictable, capital retreats from high-risk deployment and crowds into anything with a defensible category. That is why DeFi's interest-rate curves — the ones Aave and Compound publish like scripture — are not really market discovery. They are governance parameters dressed up as market discovery, set by vote, not by supply and demand. AI faces the same fiction: its compliance costs will not be discovered by "the market." They will be set by whichever court hands down the first punitive verdict.
The tail risk is what kills you. Administrative fines are survivable. A consumer class action against a deployed model that misled a million users is not. Multiply that across 50 state attorneys general and parallel EU enforcement, and you get a compliance bill that scales with your user base but arrives without warning.
Then there is geography. When penalties are set case by case, plaintiffs shop for the courtroom most likely to punish. A model deployed globally gets sued where liability is strictest and damages are highest. The AI supply chain — foundation model, fine-tuner, deployer, end user — has no clean liability boundary, so claims funnel upstream toward the deepest pocket. That is not a bug in the system. It is the system. The same funnel has been pulling capital out of crypto's long tail for years.
The hidden cost is a tax on uncertainty. Large firms absorb legal risk through scale — in-house counsel, insurance, political capital. Small teams cannot diversify a lawsuit. So the rule vacuum does not punish bad actors; it punishes the small ones, and it quietly concentrates the market into a handful of players who can survive a decade of litigation. The industry that survives is not the most innovative. It is the one that can afford the legal war.
The consensus read is that AI regulation is a tech story. It is not. It is a capital-formation story, and it is running the same script crypto already ran.
Everyone assumes clear AI rules will arrive and settle the question. They will not, because the system is designed to produce settlements, not rulings. Class actions end in confidential payouts. Consent decrees end in compliance monitors. The standard never gets litigated to a verdict, so the uncertainty never resolves. We don't get clarity from legislation; we get clarity from exhaustion, and the market prices that exhaustion long before the lawyers do.
Crypto's version of this was regulation by enforcement — no rule, just the next subpoena. AI is about to learn the same lesson from the other direction: no gatekeeper means no safe harbor, and no safe harbor means your legal bill is your moat. This is the bull market's blind spot. Everyone is auditing tokenomics and ignoring legal architecture, when the legal architecture is the thing that decides which projects still exist after the cycle turns.
So the trade is not AI stocks versus crypto. It is lawyers versus everyone. Watch the first major AI product-liability verdict the way you once watched the first token enforcement action — not for the defendant, but for the precedent. When that verdict lands, the capital that flees AI lands somewhere. Some of it lands here.
Speed wins the trade, discipline keeps the profit. Position before the precedent, not after it.