The draft ran four pages and promised almost nothing.
That was my first thought when the Crypto Briefing alert hit my feed at 6:47 a.m. Rome time. Trump had unveiled a "voluntary AI accord" โ a phrase that does an enormous amount of work for how little it actually commits to โ while lawmakers on Capitol Hill kept pushing for what they insistently brand "real oversight." No signatures with teeth. No enforcement mechanism. No stated penalty for a company that simply walks away from the table. Just a handshake, wrapped in policy language, announced with the confidence of a man who has never met a binding commitment he could not renegotiate.
I have covered this exact rhetorical move before. Not in AI policy โ in token whitepapers. The pattern never changes. You call something "voluntary" because "unregulated" polls badly. You call it a "framework" because "no rules at all" sounds like negligence. And you announce it fast, before the fine print can circulate, because in this business the headline is the product and the product is attention.
For anyone who lived through 2017, the muscle memory should be firing right now. From ICO hype to on-chain truth, we have watched this movie before โ and we know how the third act tends to end.
Here is why a crypto desk is even covering an AI governance story, and why that should tell you something before you read a single line of the accord.
The connective tissue is one appointment. David Sacks, the so-called "AI and crypto czar," holds both portfolios simultaneously. That is not a staffing accident. It is a philosophy statement. The same administration that has spent two years dismantling crypto enforcement โ pulling back cases, softening the posture of agencies that treated every token as a security through 2023 โ is now applying the identical template to artificial intelligence. Light-touch governance. Industry-friendly framing. Voluntary over mandatory. The vocabulary is interchangeable because the worldview underneath is.
So when a crypto vertical reports on an AI accord, you are not reading neutral policy coverage. You are reading a story filtered through a lens that has argued for years that permissive regulation is the correct default. That is not automatically wrong โ but you need to know the water you are swimming in.
To understand the stakes, remember what the alternative looked like. Under the previous administration, Executive Order 14110 set reporting thresholds for frontier models โ compute cutoffs that forced labs to disclose training runs above a certain scale. That order is gone. Revoked. What replaced it is not a statute, not a rule, not an enforcement regime. It is a voluntary accord and a promise that Congress might, someday, do something harder.
Meanwhile the actual hard law is being written at the state level. Colorado has passed AI legislation. California has been fighting over its own framework. Texas is moving. So we have a genuinely strange configuration: a federal government that prefers handshakes, and a patchwork of states that prefer statutes. Any company building at scale now faces the worst of both worlds โ no federal clarity to preempt the chaos, and a growing stack of state rules to satisfy one at a time.
This is where my audit instincts fire. When I spent 2017 tearing through ERC-20 whitepapers, the tell was never the technology. The tell was the gap between what a project promised and what its token contract could actually enforce. Voluntary promises and on-chain constraints are different species of thing. One bends under pressure. The other executes whether you like it or not.
A voluntary accord is a whitepaper with a press release. It is a promise about future behavior with no mechanism to make the promise binding. And I have watched enough of those promises evaporate the moment a competitor shipped faster.
Here is the part that should genuinely worry anyone who cares about either AI safety or crypto integrity โ and it is the part almost nobody is saying out loud.
A voluntary framework does not eliminate regulation. It relocates it โ from the government to the balance sheet, and from the statute book to the moat.
Think about who can actually afford to comply with a "voluntary" standard. A frontier lab with a legal department of forty people can produce the safety reports, the model cards, the red-team documentation, the third-party attestations. It can absorb the cost of performing virtue because that cost is a rounding error against its next funding round. A twenty-person startup cannot. So "voluntary" does not mean "no rules." It means rules that only the incumbents can afford to follow โ which is functionally identical to rules that only incumbents get to write.
I have watched this dynamic play out in DeFi. When Uniswap V4 shipped hooks, it turned the DEX into programmable Lego โ genuinely elegant, genuinely powerful โ and within a year the complexity spike had scared off the vast majority of would-be developers. The ones who survived were the ones with teams big enough to absorb the learning curve. Every layer of "optional" complexity is a filter, and filters favor whoever is already large. "Voluntary" is just a friendlier word for the same gravitational pull.
So when the headlines frame this as "Trump goes easy on AI," the smarter read is the opposite. The voluntary accord is not a gift to the industry. It is a subsidy to the largest players, dressed as deregulation.
And the crypto parallel is not theoretical. It is literal. The same administration that is now blessing a voluntary AI accord spent the last two years replacing crypto enforcement with... nothing in particular. Not clarity. Not safe harbors. Not a coherent rulebook. Just a slow retreat from enforcement, plus a rhetorical invitation for the industry to police itself. That is the same move, in a different vertical, with the same beneficiary.
This is where the "regulation by enforcement" critique I have carried for years gets a strange twist. For a long time, the complaint from crypto was that regulators refused to write clear rules and instead sued their way to policy. The implicit promise was always: give us a legislature that writes rules, and everything gets better. But look closely at what is happening now. The enforcement is receding, and the rules still are not arriving. We traded a hostile regulator for an absent one. That is not the same as getting a fair one.
A voluntary AI accord is the purest expression of that trade. It is the government saying: we will not force you to do anything, and we will not protect you from anything either. It is the regulatory equivalent of a shrug, delivered with a press conference.
Now let me translate this into something a builder or an investor can actually use, because framework-speak is cheap and I want to hand you something sharp.
The immediate impact is asymmetric. Low-regulatory-sensitivity AI applications โ customer service automation, content generation, developer tooling โ get an accelerant. They were never going to be governed by a general AI rule anyway; they answer to copyright law, to liability law, to whatever their customers demand. For them, the accord changes nothing, and the absence of federal friction is pure tailwind. This is where the near-term deployment surge lands, and where I would expect the fastest commercial noise.
High-sensitivity verticals get almost nothing. Healthcare AI still answers to the FDA. Financial AI still answers to financial regulators. Legal AI still answers to the bar and to malpractice exposure. A voluntary accord from the White House does not loosen any of those chains, because those chains were never forged by the White House. The deregulation is real, but it is concentrated exactly where the money was already moving fastest โ which means it amplifies the existing trend rather than redirecting it.
And here is the second-order effect that the policy desks will underweight and the market will overprice: the crypto-AI crossover narrative. The moment the same czar owns both portfolios, every AI-agent token, every "decentralized AI" pitch, every project with "autonomous" in its name gets a fresh coat of policy-approved paint. Some of it will be real. Much of it will be narrative arbitrage โ founders repackaging a compute wrapper as an AI protocol and riding the deregulation wave. Capturing the fleeting spirit of the herd is easy in weeks like this. Distinguishing the durable from the decorative is the whole job.
There is also an international dimension the crypto crowd keeps ignoring. If the United States goes soft on AI governance while the European Union hardens its risk-tiered AI Act, the standards war flips. Historically, Brussels exported its strictness โ GDPR became the global default because everyone built to it. A permissive America does not erase that gravity; it hands the EU a clearer runway to write the rules the world eventually copies. For founders planning a token launch or an AI product with global reach, the compliance bar is not set in Washington. It is set wherever the strictest meaningful jurisdiction decides to draw it.
Now for the angle I think almost everyone is missing โ including the lawmakers loudest about "real oversight."
The assumption baked into the coverage is that the voluntary accord is the weak option and mandatory oversight is the strong one. But strength and strictness are not the same thing, and the thing markets actually punish is not strictness. It is uncertainty.
A clear, strict rule is more investable than a vague, permissive one. Ask anyone who has tried to build under a regime where the goalposts move every quarter. You can plan around a hard constraint. You cannot plan around a vacuum that might fill itself at any moment. The voluntary accord's real defect is not that it is soft โ it is that it is unstable. Congress is already pushing back. The moment a real AI incident lands on the front page, the same lawmakers demanding oversight will have their mandate handed to them, and the rules could snap from voluntary to draconian in a single news cycle.
That is the scenario I would actually underwrite. Not "no regulation forever" โ but "light regulation now, followed by a hard reversal the first time something goes badly wrong." If you have watched crypto long enough, you know the strongest crackdowns never arrive on schedule. They arrive after the disaster. FTX did not get regulated on a calendar; it got regulated in the two weeks after November 2022, when the political cost of doing nothing spiked. I spent that bear market hosting recovery dinners in Rome, trading notes with people who had seen the warning signs early, and the lesson that stuck was simple: the market does not regulate the risk. It regulates the memory of the risk.
So the correct posture toward this accord is not relief. It is scenario-building. Build your internal safety bar above whatever the voluntary standard requires, precisely because the standard is likely to be replaced by something far stricter at the worst possible moment. The companies that survive regime changes are the ones already compliant with the regime that does not exist yet.
And a final, uncomfortable point for my own side of the aisle. The accord's supporters have a real argument the critical coverage keeps skipping: unilateral American regulation of AI may simply not work, because models do not respect borders. If Washington ties its own labs in knots while everyone else races ahead, the constraint exports the industry rather than the safety. I think that argument is doing more work than it can bear โ but pretending it does not exist is how you end up blindsided. Scanning the noise for the signal means reading the arguments you disagree with more carefully than the ones you don't.
The accord is not the story. The vacuum is.
Watch three things. First, whether any real enforcement lever hides inside the "voluntary" language โ federal procurement is the one that bites, because a government that will not regulate you can still refuse to buy from you. Second, whether Congress converts rhetoric into a bill, and how fast. Third, whether the first serious AI incident arrives before or after the rules do. If it arrives first, everything you just read about light-touch governance gets rewritten in a weekend.
The cheetah does not wait for the herd to decide where it is going. It watches which way the grass moves first. Right now the grass is moving toward permissive, and the wind is already shifting. Speed meets substance in the void โ and the void, this time, has a press release.