The government was hours away from signing a rule that would have required AI companies to let safety experts review their systems before releasing them to the public, specifically to protect hospitals and banks. The signing ceremony was scheduled. The rule was drafted.
Then three of the most powerful men in the AI industry communicated with the administration. Elon Musk. Mark Zuckerberg. David Sacks.
The rule was canceled.
The official reason: we can not let China get ahead.
China had nothing to do with that rule. The rule only applied to American companies, operating in America, deploying systems in America. Canceling it did not slow Chinese AI development by a single second. Canceling it removed a safety review process for systems being deployed in American hospitals and banks, systems that will make decisions about your medical care and your money.
China was not in that phone call. Three American billionaires were.
This is what regulatory capture looks like in the AI era. Not a dramatic scandal. Not a bribe. Not a conspiracy. Just three phone calls, and a safety rule dead before most people knew it existed.
THE LOBBYING NUMBERS
The lobbying numbers make the pattern explicit. By the first quarter of 2026, Anthropic was spending $1.6 million per quarter on federal lobbying. OpenAI was spending $1.5 million. Meta was spending $7.1 million per quarter. The seven largest technology companies spent over $50 million on AI lobbying in nine months of 2025.
In the same period, the entire AI safety research field, all of it, worldwide, received less in grants than OpenAI alone spent on lobbyists.
The companies warning most loudly about AI’s catastrophic dangers are simultaneously spending tens of millions of dollars ensuring that the regulations those warnings imply are never implemented. This is not a contradiction. It is a strategy. The fear generates goodwill. The lobbying ensures that goodwill never becomes binding law.
THE CONFLICT AT THE TOP
David Sacks was appointed by the Trump administration as America’s AI and Crypto Czar, the government official most directly responsible for shaping federal AI policy. He is simultaneously a general partner at Craft Ventures, a venture capital firm with an active portfolio of AI startup investments.
This is not a historical fact. This is the current arrangement. The person setting national AI policy holds financial positions in the industry he is setting policy for. His explicit stated position is that AI regulation should be minimally burdensome, a position that, by a remarkable coincidence, serves the commercial interests of the companies in his portfolio.
He has publicly accused Anthropic of running a sophisticated regulatory capture strategy based on fear-mongering. He is correct about the mechanism. He is also engaged in the inverse version of the same strategy: using a government position to eliminate regulations that would inconvenience his portfolio companies while framing deregulation as American competitiveness against China.
Both Sacks and Anthropic are engaged in regulatory capture. They are capturing in opposite directions. Anthropic wants regulations it helped design. Sacks wants no regulations at all. But both are pursuing outcomes that serve specific financial interests at the expense of democratic governance.
THE CHINA ARGUMENT AND WHY IT DOES NOT HOLD
The single most powerful rhetorical weapon in the AI industry’s regulatory capture arsenal is the China argument. We can not let China beat us, or its variants, appears in virtually every argument against AI safety legislation, environmental review, community consultation, or any other mechanism that might slow American AI development.
First: American AI safety regulations do not affect what China builds. This is a simple empirical fact. A requirement that American AI companies submit to pre-deployment safety review has zero effect on the research output of Chinese AI labs.
Second: The specific regulations being opposed are rarely about AI capability. The executive order that was canceled after phone calls from Musk, Zuckerberg, and Sacks would have required AI companies to show their systems to federal safety experts before releasing them, to protect banks and hospitals. This is not a capability constraint. It is a review process.
Third: The countries that regulate AI most carefully are not losing the competition. The European Union’s AI Act is the most comprehensive AI regulation in the world. European AI companies are not visibly being outcompeted by their American counterparts because of it.
THE THREE PHONE CALLS
Three phone calls. The precise mechanics are not fully documented. What is documented is the outcome: an executive order designed to protect American hospitals and banks was canceled, hours before signing, after communication between the President and three of the most powerful figures in the AI industry.
This happened in a democracy. It happened to a safety measure that was never publicly debated, never put to a vote, never given the opportunity to generate the public support that a regulation protecting hospital systems from AI cyberattacks would likely have received. It was killed before it could be seen.
The people who made those calls are not villains in the cinematic sense. They are rational actors operating within an incentive structure that rewards the elimination of regulation and punishes safety investment. The structure is the problem.
The question is whether the people who did not make those calls, the rest of us, are willing to build a different structure. One in which the phone calls go differently. In which the safety rule gets signed. In which the community gets consulted. In which the water stays in the aquifer.
That structure does not build itself. It is built by people who understand what is happening and decide that it matters enough to act.
Sources: Axios · NPR · Federal lobbying disclosures Q1 2026 · Craft Ventures public filings · Executive order reporting May 2026
Cricketpocalypse is an independent channel. No corporate funding. No AI company money. Just the facts and the bugs.